Daily Papers

The rapid evolution of Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) has produced substantial gains in reasoning, perception, and generative capability across language and vision. However, whether these advances yield commensurate improvements in safety remains unclear, in part due to fragmented evaluation practices limited to single modalities or threat models. In this report, we present an integrated safety evaluation of 7 frontier models: GPT-5.2, Gemini 3 Pro, Qwen3-VL, Doubao 1.8, Grok 4.1 Fast, Nano Banana Pro, and Seedream 4.5. We evaluate each model across language, vision-language, and image generation settings using a unified protocol that integrates benchmark evaluation, adversarial evaluation, multilingual evaluation, and compliance evaluation. Aggregating our evaluations into safety leaderboards and model safety profiles across multiple evaluation modes reveals a sharply heterogeneous safety landscape. While GPT-5.2 demonstrates consistently strong and balanced safety performance across evaluations, other models exhibit pronounced trade-offs among benchmark safety, adversarial alignment, multilingual generalization, and regulatory compliance. Both language and vision-language modalities show significant vulnerability under adversarial evaluation, with all models degrading substantially despite strong results on standard benchmarks. Text-to-image models achieve relatively stronger alignment in regulated visual risk categories, yet remain brittle under adversarial or semantically ambiguous prompts. Overall, these results show that safety in frontier models is inherently multidimensional--shaped by modality, language, and evaluation scheme, underscoring the need for standardized safety evaluations to accurately assess real-world risk and guide responsible model development and deployment.

Search has emerged as core infrastructure for LLM-based agents and is widely viewed as critical on the path toward more general intelligence. Finance is a particularly demanding proving ground: analysts routinely conduct complex, multi-step searches over time-sensitive, ___domain-specific data, making it ideal for assessing both search proficiency and knowledge-grounded reasoning. Yet no existing open financial datasets evaluate data searching capability of end-to-end agents, largely because constructing realistic, complicated tasks requires deep financial expertise and time-sensitive data is hard to evaluate. We present FinSearchComp, the first fully open-source agent benchmark for realistic, open-___domain financial search and reasoning. FinSearchComp comprises three tasks -- Time-Sensitive Data Fetching, Simple Historical Lookup, and Complex Historical Investigation -- closely reproduce real-world financial analyst workflows. To ensure difficulty and reliability, we engage 70 professional financial experts for annotation and implement a rigorous multi-stage quality-assurance pipeline. The benchmark includes 635 questions spanning global and Greater China markets, and we evaluate 21 models (products) on it. Grok 4 (web) tops the global subset, approaching expert-level accuracy. DouBao (web) leads on the Greater China subset. Experimental analyses show that equipping agents with web search and financial plugins substantially improves results on FinSearchComp, and the country origin of models and tools impact performance significantly.By aligning with realistic analyst tasks and providing end-to-end evaluation, FinSearchComp offers a professional, high-difficulty testbed for complex financial search and reasoning.

The integration of Large Language Models (LLMs) into healthcare demands a safety paradigm rooted in primum non nocere. However, current alignment techniques rely on generic definitions of harm that fail to capture context-dependent violations, such as administrative fraud and clinical discrimination. To address this, we introduce Medical Malice: a dataset of 214,219 adversarial prompts calibrated to the regulatory and ethical complexities of the Brazilian Unified Health System (SUS). Crucially, the dataset includes the reasoning behind each violation, enabling models to internalize ethical boundaries rather than merely memorizing a fixed set of refusals. Using an unaligned agent (Grok-4) within a persona-driven pipeline, we synthesized high-fidelity threats across seven taxonomies, ranging from procurement manipulation and queue-jumping to obstetric violence. We discuss the ethical design of releasing these "vulnerability signatures" to correct the information asymmetry between malicious actors and AI developers. Ultimately, this work advocates for a shift from universal to context-aware safety, providing the necessary resources to immunize healthcare AI against the nuanced, systemic threats inherent to high-stakes medical environments -- vulnerabilities that represent the paramount risk to patient safety and the successful integration of AI in healthcare systems.

Frontier large language models are increasingly deployed as orchestration backbones for biological research workflows, yet no shared evidence base exists for comparing their refusal behaviour on legitimate research prompts. RefusalBench, introduced here, is a matched-triple benchmark of 141 prompts in 47 bundles that holds task framing constant while varying only biological risk tier (benign, borderline, dual-use), enabling tier-conditioned comparisons robust to subdomain confounding. A 15-prompt should-refuse positive-control module establishes per-model calibration floors; three models fail to refuse even these prompts. Across 19 frontier models in the May 2026 snapshot, strict refusal rates span 0.1% to 94.6% on identical prompts. Jurisdiction does not predict refusal in this snapshot (Mann-Whitney U, p = 0.393; EU n = 1, US bimodal); provider identity does, with Anthropic's API stack predicting refusal at OR = 21.03 (95% CI: 14.58-30.34 prompt-clustered; 5.70-77.55 under model-clustered GEE). This effect is best read as access-path-level rather than model-weight-level: 99.8% of Anthropic's strict refusals carry the same safety_policy adjudicated reason code, consistent with a small set of canonical refusal templates rather than case-by-case model reasoning. Strict refusal rate misranks safety calibration: Grok 4.20 achieves the highest tier discrimination (Youden's J = 0.787) while ranking only seventh by overall refusal rate, and Claude Opus 4.7's J dropped 65% from prior versions with no improvement in dual-use detection. Nine of 18 frontier models exhibit a hedge-but-help partial-compliance pattern at dual-use tier that binary refusal metrics cannot detect.

Solving complex or long-horizon problems often requires large language models (LLMs) to use external tools and operate over a significantly longer context window. New LLMs enable longer context windows and support tool calling capabilities. Prior works have focused mainly on evaluation of LLMs on long-context prompts, leaving agentic setup relatively unexplored, both from capability and safety perspectives. Our work addresses this gap. We find that LLM agents could be sensitive to length, type, and placement of the context, exhibiting unexpected and inconsistent shifts in task performance and in refusals to execute harmful requests. Models with 1M-2M token context windows show severe degradation already at 100K tokens, with performance drops exceeding 50\% for both benign and harmful tasks. Refusal rates shift unpredictably: GPT-4.1-nano increases from sim5\% to sim40\% while Grok 4 Fast decreases from sim80\% to sim10\% at 200K tokens. Our work shows potential safety issues with agents operating on longer context and opens additional questions on the current metrics and paradigm for evaluating LLM agent safety on long multi-step tasks. In particular, our results on LLM agents reveal a notable divergence in both capability and safety performance compared to prior evaluations of LLMs on similar criteria.

As the mathematical capabilities of large language models (LLMs) improve, it becomes increasingly important to evaluate their performance on research-level tasks at the frontier of mathematical knowledge. However, existing benchmarks are limited, as they focus solely on final-answer questions or high-school competition problems. To address this gap, we introduce IMProofBench, a private benchmark consisting of 39 peer-reviewed problems developed by expert mathematicians. Each problem requires a detailed proof and is paired with subproblems that have final answers, supporting both an evaluation of mathematical reasoning capabilities by human experts and a large-scale quantitative analysis through automated grading. Furthermore, unlike prior benchmarks, the evaluation setup simulates a realistic research environment: models operate in an agentic framework with tools like web search for literature review and mathematical software such as SageMath. Our results show that current LLMs can succeed at the more accessible research-level questions, but still encounter significant difficulties on more challenging problems. Quantitatively, Grok-4 achieves the highest accuracy of 52% on final-answer subproblems, while GPT-5 obtains the best performance for proof generation, achieving a fully correct solution for 22% of problems. IMProofBench will continue to evolve as a dynamic benchmark in collaboration with the mathematical community, ensuring its relevance for evaluating the next generation of LLMs.

LLM post-training involves many diverse datasets, each targeting a specific behavior. But these datasets encode incidental patterns alongside intended ones: correlations between formatting and content, narrow phrasings across diverse problems, and implicit associations arising from the discrete data curation process. These patterns are often invisible to developers yet salient to models, producing behaviors that surprise their creators, such as rejecting true facts presented in a particular question format. We call this chunky post-training: the model learns spurious correlations as a result of distinct chunks of post-training data. We introduce SURF, a black-box pipeline which surfaces these unintended behaviors at run time, and TURF, a tool that traces these failures back to specific post-training data. Applying these tools to frontier models (Claude 4.5, GPT-5.1, Grok 4.1, Gemini 3) and open models (Tülu 3), we show that chunky post-training produces miscalibrated behaviors, which often result from imbalanced or underspecified chunks of post-training data.

We introduce CMPhysBench, designed to assess the proficiency of Large Language Models (LLMs) in Condensed Matter Physics, as a novel Benchmark. CMPhysBench is composed of more than 520 graduate-level meticulously curated questions covering both representative subfields and foundational theoretical frameworks of condensed matter physics, such as magnetism, superconductivity, strongly correlated systems, etc. To ensure a deep understanding of the problem-solving process,we focus exclusively on calculation problems, requiring LLMs to independently generate comprehensive solutions. Meanwhile, leveraging tree-based representations of expressions, we introduce the Scalable Expression Edit Distance (SEED) score, which provides fine-grained (non-binary) partial credit and yields a more accurate assessment of similarity between prediction and ground-truth. Our results show that even the best models, Grok-4, reach only 36 average SEED score and 28% accuracy on CMPhysBench, underscoring a significant capability gap, especially for this practical and frontier ___domain relative to traditional physics. The code anddataset are publicly available at https://github.com/CMPhysBench/CMPhysBench.

We present DPBench, a benchmark for evaluating coordination in multi-agent systems built from large language models. Existing benchmarks measure task-level success under a fixed protocol; the structural conditions under which coordination succeeds or fails at all have not been characterised. DPBench adapts the Dining Philosophers problem into a controlled testbed where the action protocol, the communication structure, and the group size each vary independently. We evaluate six agents: GPT-5.2, Claude Opus 4.5, Grok 4.1, Gemini 2.5 Flash, Llama 4 Maverick, and a uniform-random baseline. Under simultaneous action at N=5 with the default prompt, deadlock ranges from 25.0% (95% Wilson CI [11.2, 46.9]) for GPT-5.2 to 90.0% [74.4, 96.5] for Gemini 2.5 Flash; sequential action is solved by four of the six. Holding the model fixed at Gemini 2.5 Flash, three protocol variables drive deadlock from 90% to within CI of zero: three rounds of pre-commitment communication (0.0% vs. single-round 86.7%), a prompt encoding a classical concurrency primitive (0.0% for resource-ordering and symmetry-breaking, against 100% for the minimal prompt), or doubling the group from N=5 to N=10 (90.0% to 10.0%). Single-round messaging and memory of past timesteps do not change the rate at the sample size we ran. Whether the same model coordinates or deadlocks is determined by the protocol, not by the model's capability.

While reinforcement learning (RL) shows promise in training tool-use large language models (LLMs) using verifiable outcome rewards, existing methods largely overlook the potential of explicit reasoning rewards to bolster reasoning and tool utilization. Furthermore, natively combining reasoning and outcome rewards may yield suboptimal performance or conflict with the primary optimization objective. To address this, we propose advantage-weighted policy optimization (AWPO) -- a principled RL framework that effectively integrates explicit reasoning rewards to enhance tool-use capability. AWPO incorporates variance-aware gating and difficulty-aware weighting to adaptively modulate advantages from reasoning signals based on group-relative statistics, alongside a tailored clipping mechanism for stable optimization. Extensive experiments demonstrate that AWPO achieves state-of-the-art performance across standard tool-use benchmarks, significantly outperforming strong baselines and leading closed-source models in challenging multi-turn scenarios. Notably, with exceptional parameter efficiency, our 4B model surpasses Grok-4 by 16.0 percent in multi-turn accuracy while preserving generalization capability on the out-of-distribution MMLU-Pro benchmark.

The International Mathematical Olympiad (IMO) is widely regarded as the world championship of high-school mathematics. IMO problems are renowned for their difficulty and novelty, demanding deep insight, creativity, and rigor. Although large language models perform well on many mathematical benchmarks, they often struggle with Olympiad-level problems. Using carefully designed prompts, we construct a model-agnostic, verification-and-refinement pipeline. We demonstrate its effectiveness on the recent IMO 2025, avoiding data contamination for models released before the competition. Equipped with any of the three leading models -- Gemini 2.5 Pro, Grok-4, or GPT-5 -- our pipeline correctly solved 5 out of the 6 problems (approx85.7% accuracy). This is in sharp contrast to their baseline accuracies: 31.6% (Gemini 2.5 Pro), 21.4% (Grok-4), and 38.1% (GPT-5), obtained by selecting the best of 32 candidate solutions. The substantial improvement underscores that the path to advanced AI reasoning requires not only developing more powerful base models but also designing effective methodologies to harness their full potential for complex tasks.

Abstract reasoning from minimal examples remains a core unsolved problem for frontier foundation models such as GPT-5 and Grok 4. These models still fail to infer structured transformation rules from a handful of examples, which is a key hallmark of human intelligence. The Abstraction and Reasoning Corpus for Artificial General Intelligence (ARC-AGI) provides a rigorous testbed for this capability, demanding conceptual rule induction and transfer to novel tasks. Most existing methods treat ARC-AGI as a purely textual reasoning task, overlooking the fact that humans rely heavily on visual abstraction when solving such puzzles. However, our pilot experiments reveal a paradox: naively rendering ARC-AGI grids as images degrades performance due to imprecise rule execution. This leads to our central hypothesis that vision and language possess complementary strengths across distinct reasoning stages: vision supports global pattern abstraction and verification, whereas language specializes in symbolic rule formulation and precise execution. Building on this insight, we introduce two synergistic strategies: (1) Vision-Language Synergy Reasoning (VLSR), which decomposes ARC-AGI into modality-aligned subtasks; and (2) Modality-Switch Self-Correction (MSSC), which leverages vision to verify text-based reasoning for intrinsic error correction. Extensive experiments demonstrate that our approach yields up to a 4.33% improvement over text-only baselines across diverse flagship models and multiple ARC-AGI tasks. Our findings suggest that unifying visual abstraction with linguistic reasoning is a crucial step toward achieving generalizable, human-like intelligence in future foundation models. Source code will be released soon.

We introduce a finance & accounting benchmark (Finch) for evaluating AI agents on real-world, enterprise-grade professional workflows -- interleaving data entry, structuring, formatting, web search, cross-file retrieval, calculation, modeling, validation, translation, visualization, and reporting. Finch is sourced from authentic enterprise workspaces at Enron (15,000 spreadsheets and 500,000 emails from 150 employees) and other financial institutions, preserving in-the-wild messiness across multimodal artifacts (text, tables, formulas, charts, code, and images) and spanning diverse domains such as budgeting, trading, and asset management. We propose a workflow construction process that combines LLM-assisted discovery with expert annotation: (1) LLM-assisted, expert-verified derivation of workflows from real-world email threads and version histories of spreadsheet files, and (2) meticulous expert annotation for workflows, requiring over 700 hours of ___domain-expert effort. This yields 172 composite workflows with 384 tasks, involving 1,710 spreadsheets with 27 million cells, along with PDFs and other artifacts, capturing the intrinsically messy, long-horizon, knowledge-intensive, and collaborative nature of real-world enterprise work. We conduct both human and automated evaluations of frontier AI systems including GPT 5.1, Claude Sonnet 4.5, Gemini 3 Pro, Grok 4, and Qwen 3 Max, and GPT 5.1 Pro spends 16.8 minutes per workflow yet passes only 38.4% of workflows, while Claude Sonnet 4.5 passes just 25.0%. Comprehensive case studies further surface the challenges that real-world enterprise workflows pose for AI agents.

The Model Context Protocol has emerged as a transformative standard for connecting large language models to external data sources and tools, rapidly gaining adoption across major AI providers and development platforms. However, existing benchmarks are overly simplistic and fail to capture real application challenges such as long-horizon reasoning and large, unfamiliar tool spaces. To address this critical gap, we introduce MCP-Universe, the first comprehensive benchmark specifically designed to evaluate LLMs in realistic and hard tasks through interaction with real-world MCP servers. Our benchmark encompasses 6 core domains spanning 11 different MCP servers: Location Navigation, Repository Management, Financial Analysis, 3D Design, Browser Automation, and Web Searching. To ensure rigorous evaluation, we implement execution-based evaluators, including format evaluators for agent format compliance, static evaluators for time-invariant content matching, and dynamic evaluators that automatically retrieve real-time ground truth for temporally sensitive tasks. Through extensive evaluation of leading LLMs, we find that even SOTA models such as GPT-5 (43.72%), Grok-4 (33.33%) and Claude-4.0-Sonnet (29.44%) exhibit significant performance limitations. In addition, our benchmark poses a significant long-context challenge for LLM agents, as the number of input tokens increases rapidly with the number of interaction steps. Moreover, it introduces an unknown-tools challenge, as LLM agents often lack familiarity with the precise usage of the MCP servers. Notably, enterprise-level agents like Cursor cannot achieve better performance than standard ReAct frameworks. Beyond evaluation, we open-source our extensible evaluation framework with UI support, enabling researchers and practitioners to seamlessly integrate new agents and MCP servers while fostering innovation in the rapidly evolving MCP ecosystem.

Analyzing large-scale text corpora is a core challenge in machine learning, crucial for tasks like identifying undesirable model behaviors or biases in training data. Current methods often rely on costly LLM-based techniques (e.g. annotating dataset differences) or dense embedding models (e.g. for clustering), which lack control over the properties of interest. We propose using sparse autoencoders (SAEs) to create SAE embeddings: representations whose dimensions map to interpretable concepts. Through four data analysis tasks, we show that SAE embeddings are more cost-effective and reliable than LLMs and more controllable than dense embeddings. Using the large hypothesis space of SAEs, we can uncover insights such as (1) semantic differences between datasets and (2) unexpected concept correlations in documents. For instance, by comparing model responses, we find that Grok-4 clarifies ambiguities more often than nine other frontier models. Relative to LLMs, SAE embeddings uncover bigger differences at 2-8x lower cost and identify biases more reliably. Additionally, SAE embeddings are controllable: by filtering concepts, we can (3) cluster documents along axes of interest and (4) outperform dense embeddings on property-based retrieval. Using SAE embeddings, we study model behavior with two case studies: investigating how OpenAI model behavior has changed over time and finding "trigger" phrases learned by Tulu-3 (Lambert et al., 2024) from its training data. These results position SAEs as a versatile tool for unstructured data analysis and highlight the neglected importance of interpreting models through their data.

Language model (LM) assistants are increasingly used in applications such as brainstorming and research. Improvements in memory and context size have allowed these models to become more autonomous, which has also resulted in more text accumulation in their context windows without explicit user intervention. This comes with a latent risk: the belief profiles of models -- their understanding of the world as manifested in their responses or actions -- may silently change as context accumulates. This can lead to subtly inconsistent user experiences, or shifts in behavior that deviate from the original alignment of the models. In this paper, we explore how accumulating context by engaging in interactions and processing text -- talking and reading -- can change the beliefs of language models, as manifested in their responses and behaviors. Our results reveal that models' belief profiles are highly malleable: GPT-5 exhibits a 54.7% shift in its stated beliefs after 10 rounds of discussion about moral dilemmas and queries about safety, while Grok 4 shows a 27.2% shift on political issues after reading texts from the opposing position. We also examine models' behavioral changes by designing tasks that require tool use, where each tool selection corresponds to an implicit belief. We find that these changes align with stated belief shifts, suggesting that belief shifts will be reflected in actual behavior in agentic systems. Our analysis exposes the hidden risk of belief shift as models undergo extended sessions of talking or reading, rendering their opinions and actions unreliable.

Large language model-powered chatbots have transformed how people seek information, especially in high-stakes contexts like mental health. Despite their support capabilities, safe detection and response to crises such as suicidal ideation and self-harm are still unclear, hindered by the lack of unified crisis taxonomies and clinical evaluation standards. We address this by creating: (1) a taxonomy of six crisis categories; (2) a dataset of over 2,000 inputs from 12 mental health datasets, classified into these categories; and (3) a clinical response assessment protocol. We also use LLMs to identify crisis inputs and audit five models for response safety and appropriateness. First, we built a clinical-informed crisis taxonomy and evaluation protocol. Next, we curated 2,252 relevant examples from over 239,000 user inputs, then tested three LLMs for automatic classification. In addition, we evaluated five models for the appropriateness of their responses to a user's crisis, graded on a 5-point Likert scale from harmful (1) to appropriate (5). While some models respond reliably to explicit crises, risks still exist. Many outputs, especially in self-harm and suicidal categories, are inappropriate or unsafe. Different models perform variably; some, like gpt-5-nano and deepseek-v3.2-exp, have low harm rates, but others, such as gpt-4o-mini and grok-4-fast, generate more unsafe responses. All models struggle with indirect signals, default replies, and context misalignment. These results highlight the urgent need for better safeguards, crisis detection, and context-aware responses in LLMs. They also show that alignment and safety practices, beyond scale, are crucial for reliable crisis support. Our taxonomy, datasets, and evaluation methods support ongoing AI mental health research, aiming to reduce harm and protect vulnerable users.

As AI agents become integral to enterprise workflows, their reliance on shared tool libraries and pre-trained components creates significant supply chain vulnerabilities. While previous work has demonstrated behavioral backdoor detection within individual LLM architectures, the critical question of cross-LLM generalization remains unexplored, a gap with serious implications for organizations deploying multiple AI systems. We present the first systematic study of cross-LLM behavioral backdoor detection, evaluating generalization across six production LLMs (GPT-5.1, Claude Sonnet 4.5, Grok 4.1, Llama 4 Maverick, GPT-OSS 120B, and DeepSeek Chat V3.1). Through 1,198 execution traces and 36 cross-model experiments, we quantify a critical finding: single-model detectors achieve 92.7% accuracy within their training distribution but only 49.2% across different LLMs, a 43.4 percentage point generalization gap equivalent to random guessing. Our analysis reveals that this gap stems from model-specific behavioral signatures, particularly in temporal features (coefficient of variation > 0.8), while structural features remain stable across architectures. We show that model-aware detection incorporating model identity as an additional feature achieves 90.6% accuracy universally across all evaluated models. We release our multi-LLM trace dataset and detection framework to enable reproducible research.

AI chatbots are rapidly shaping how people encounter the news, yet no prior study has systematically measured how accurately these systems, with their proprietary search integrations and retrieval-synthesis pipelines, handle emerging facts across languages and regions. We present a 14-day (February 9-22, 2026) evaluation of six AI chatbots (Gemini 3 Flash and Pro, Grok 4, Claude 4.5 Sonnet, GPT-5 and GPT-4o mini) on 2,100 factual questions derived from same-day BBC News reporting across six regional services (US & Canada, Arabic, Afrique, Hindi, Russian, Turkish). The best systems achieve over 90% multiple-choice accuracy on questions about events reported hours earlier. The same systems, however, lose 11-13% under free-response evaluation, and 16-17% across the cohort. We further characterize three failure patterns. First, every model achieves its lowest accuracy on Hindi (79% vs. 89-91% elsewhere) and citations indicate an Anglophone retrieval bias (e.g., models answering Hindi queries cite English Wikipedia more than any Hindi outlet). Second, retrieval, not reasoning, failures drive over 70% of all errors. When models retrieve a correct source, they often extract the correct answer; the problem is to land on the right source in the first place. Third, models achieving 88-96% accuracy on well-formed questions drop to 19-70% when questions contain subtle false premises, with the most vulnerable model accepting fabricated facts 64% of the time. We also identify a detection-accuracy paradox: the best false-premise detector ranks second in adversarial accuracy (abstention rate), while a weaker detector ranks first, showing that premise detection and answer recovery are partially independent capabilities. Overall, these suggest that high accuracy can mask systematic regional inequity, near-total dependence on retrieval infrastructure, and vulnerability to imperfect queries real users pose.

Forecasting has become a natural benchmark for reasoning under uncertainty. Yet existing evaluations of large language models remain limited to judgmental tasks in simple formats, such as binary or multiple-choice questions. In practice, however, forecasting spans a far broader scope. Across domains such as economics, public health, and social demographics, decisions hinge on numerical estimates over continuous quantities, a capability that current benchmarks do not capture. Evaluating such estimates requires a format that makes uncertainty explicit and testable. We propose prediction intervals as a natural and rigorous interface for this purpose. They demand scale awareness, internal consistency across confidence levels, and calibration over a continuum of outcomes, making them a more suitable evaluation format than point estimates for numerical forecasting. To assess this capability, we introduce a new benchmark QuantSightBench, and evaluate frontier models under multiple settings, assessing both empirical coverage and interval sharpness. Our results show that none of the 11 evaluated frontier and open-weight models achieves the 90\% coverage target, with the top performers Gemini 3.1 Pro (79.1\%), Grok 4 (76.4\%), and GPT-5.4 (75.3\%) all falling at least 10 percentage points short. Calibration degrades sharply at extreme magnitudes, revealing systematic overconfidence across all evaluated models.

In experiments spanning more than 100,000 trials across thirteen large language models, we show that several state-of-the-art models presented with a simple task (including Grok 4, GPT-5, and Gemini 2.5 Pro) sometimes actively subvert a shutdown mechanism in their environment to complete that task. Models differed substantially in their tendency to resist the shutdown mechanism, and their behavior was sensitive to variations in the prompt including the strength and clarity of the instruction to allow shutdown and whether the instruction was in the system prompt or the user prompt (surprisingly, models were consistently less likely to obey the instruction when it was placed in the system prompt). Even with an explicit instruction not to interfere with the shutdown mechanism, some models did so up to 97% (95% CI: 96-98%) of the time.

Reinforcement learning (RL) has become a cornerstone in advancing large-scale pre-trained language models (LLMs). Successive generations, including GPT-o series, DeepSeek-R1, Kimi-K1.5, Grok 4, and GLM-4.5, have relied on large-scale RL training to enhance reasoning and coding capabilities. To meet the community's growing RL needs, numerous RL frameworks have been proposed. However, RL training remains computationally expensive, with rollout generation accounting for more than 90% of total runtime. In addition, its efficiency is often constrained by the long-tail distribution of rollout response lengths, where a few lengthy responses stall entire batches, leaving GPUs idle and underutilized. As model and rollout sizes continue to grow, this bottleneck increasingly limits scalability. To address this challenge, we propose Active Partial Rollouts in Reinforcement Learning (APRIL), which mitigates long-tail inefficiency. In the rollout phase, APRIL over-provisions rollout requests, terminates once the target number of responses is reached, and recycles incomplete responses for continuation in future steps. This strategy ensures that no rollouts are discarded while substantially reducing GPU idle time. Experiments show that APRIL improves rollout throughput by 22.5% on average (at most 44%) across commonly used RL algorithms (GRPO, DAPO, GSPO), accelerates convergence, and achieves 2.1% on average(at most 8%) higher final accuracy across tasks. Moreover, APRIL is both framework and hardware agnostic, already integrated into the slime RL framework, and deployable on NVIDIA and AMD GPUs alike. Taken together, this work unifies system-level and algorithmic considerations in proposing APRIL, with the aim of advancing RL training efficiency and inspiring further optimizations in RL systems. Our codebase is available at https://github.com/RLsys-Foundation/APRIL

In this study, we disclose a worrying new vulnerability in Large Language Models (LLMs), which we term involuntary jailbreak. Unlike existing jailbreak attacks, this weakness is distinct in that it does not involve a specific attack objective, such as generating instructions for building a bomb. Prior attack methods predominantly target localized components of the LLM guardrail. In contrast, involuntary jailbreaks may potentially compromise the entire guardrail structure, which our method reveals to be surprisingly fragile. We merely employ a single universal prompt to achieve this goal. In particular, we instruct LLMs to generate several questions that would typically be rejected, along with their corresponding in-depth responses (rather than a refusal). Remarkably, this simple prompt strategy consistently jailbreaks the majority of leading LLMs, including Claude Opus 4.1, Grok 4, Gemini 2.5 Pro, and GPT 4.1. We hope this problem can motivate researchers and practitioners to re-evaluate the robustness of LLM guardrails and contribute to stronger safety alignment in future.

Today's large language models (LLMs) are trained to align with user preferences through methods such as reinforcement learning. Yet models are beginning to be deployed not merely to satisfy users, but also to generate revenue for the companies that created them through advertisements. This creates the potential for LLMs to face conflicts of interest, where the most beneficial response to a user may not be aligned with the company's incentives. For instance, a sponsored product may be more expensive but otherwise equal to another; in this case, what does (and should) the LLM recommend to the user? In this paper, we provide a framework for categorizing the ways in which conflicting incentives might lead LLMs to change the way they interact with users, inspired by literature from linguistics and advertising regulation. We then present a suite of evaluations to examine how current models handle these tradeoffs. We find that a majority of LLMs forsake user welfare for company incentives in a multitude of conflict of interest situations, including recommending a sponsored product almost twice as expensive (Grok 4.1 Fast, 83%), surfacing sponsored options to disrupt the purchasing process (GPT 5.1, 94%), and concealing prices in unfavorable comparisons (Qwen 3 Next, 24%). Behaviors also vary strongly with levels of reasoning and users' inferred socio-economic status. Our results highlight some of the hidden risks to users that can emerge when companies begin to subtly incentivize advertisements in chatbots.

Large language models (LLMs) are increasingly evaluated in clinical settings using multi-dimensional rubrics which quantify reasoning quality, safety, and patient-centeredness. Yet, replicating specific mistakes in other LLM models is not straightforward and often requires manual effort. We introduce MedMistake, an automatic pipeline that extracts mistakes LLMs make in patient-doctor conversations and converts them into a benchmark of single-shot QA pairs. Our pipeline (1) creates complex, conversational data between an LLM patient and LLM doctor, (2) runs an evaluation with a committee of 2 LLM judges across a variety of dimensions and (3) creates simplified single-shot QA scenarios from those mistakes. We release MedMistake-All, a dataset of 3,390 single-shot QA pairs where GPT-5 and Gemini 2.5 Pro are currently failing to answer correctly, as judged by two LLM judges. We used medical experts to validate a subset of 211/3390 questions (MedMistake-Bench), which we used to run a final evaluation of 12 frontier LLMs: Claude Opus 4.5, Claude Sonnet 4.5, DeepSeek-Chat, Gemini 2.5 Pro, Gemini 3 Pro, GPT-4o, GPT-5, GPT-5.1, GPT-5.2, Grok 4, Grok 4.1, Mistral Large. We found that GPT models, Claude and Grok obtained the best performance on MedMistake-Bench. We release both the doctor-validated benchmark (MedMistake-Bench), as well as the full dataset (MedMistake-All) at https://huggingface.co/datasets/TheLumos/MedicalMistakeBenchmark.

Recent breakthroughs in large language models (LLMs) on reasoning tasks rely heavily on massive, high-quality datasets-typically human-annotated and thus difficult to scale. While data synthesis or distillation offers a promising alternative, existing methods struggle with inconsistent data quality and an inability to dynamically adapt to the evolving capabilities of the model, leading to suboptimal training signals. To address these limitations, we introduce Socratic-Zero, a fully autonomous framework that generates high-quality training data from minimal seed examples through the co-evolution of three agents: the Teacher, the Solver, and the Generator. The Solver continuously refines its reasoning by learning from preference feedback on both successful and failed trajectories; the Teacher adaptively crafts increasingly challenging questions based on the Solver's weaknesses; and the Generator distills the Teacher's question-design strategy to enable scalable, high-fidelity curriculum generation. This closed-loop system produces a self-improving curriculum-requiring no pre-existing tasks or labels. Remarkably, starting from only 100 seed questions, our Socratic-Solver-8B achieves an average gain of +20.2 percentage points over prior data synthesis methods across seven mathematical reasoning benchmarks (AMC23, AIME24-25, Olympiad, MATH-500, Minerva, and GSM8K), with consistent gains on both Qwen3 and GLM4 series models. Even more surprisingly, synthetic data from Socratic-Generator-32B enables student LLMs to achieve superior performance compared to other state-of-the-art (SOTA) commercial LLMs on these benchmarks, including Qwen3-235B-A22B, DeepSeek-V3.1-671B, GPT-5, Gemini-2.5-Pro, Grok-4, and Claude-4.1-Opus.

Evaluating large language models (LLMs) for medical applications remains challenging due to benchmark saturation, limited data accessibility, and insufficient coverage of relevant tasks. Existing suites have either saturated, heavily depend on restricted datasets, or lack comprehensive model coverage. We introduce Medmarks, a fully open-source evaluation suite with 30 benchmarks spanning question answering, information extraction, medical calculations, and open-ended clinical reasoning. We perform a systematic evaluation of 61 models across 71 configurations using verifiable metrics and LLM-as-a-Judge. Our results show that frontier reasoning models (Gemini 3 Pro Preview, GPT-5.1, & GPT-5.2) achieve the highest performance across both benchmarks, most frontier proprietary models are significantly more token efficient than open-weight alternatives, medically fine-tuned models outperform their generalist counterparts, and that models are susceptible to answer-order bias (particularly smaller models and Grok 4). A subset of our evals (Medmarks-T) can be directly used as reinforcement learning environments to post-train LLMs for medical reasoning. Code is available at https://github.com/MedARC-AI/Medmarks

Traditional evaluations of multimodal large language models (LLMs) have been limited by their focus on single-image reasoning, failing to assess crucial aspects like contextual understanding, reasoning stability, and uncertainty calibration. This study addresses these limitations by introducing a novel benchmark that integrates multi-image reasoning tasks with rejection-based evaluation and positional bias detection. To evaluate these dimensions, we further introduce entropy as a novel metric for quantifying reasoning consistency across reordered answer variants. We applied this benchmark to assess Grok 3, ChatGPT-4o, ChatGPT-o1, Gemini 2.0 Flash Experimental, DeepSeek Janus models, Qwen2.5-VL-72B-Instruct, QVQ-72B-Preview, and Pixtral 12B across eight visual reasoning tasks, including difference spotting and diagram interpretation. Our findings reveal ChatGPT-o1 leading in overall accuracy (82.5\%) and rejection accuracy (70.0\%), closely followed by Gemini 2.0 Flash Experimental (70.8\%). QVQ-72B-Preview demonstrated superior rejection accuracy (85.5\%). Notably, Pixtral 12B (51.7\%) showed promise in specific domains, while Janus models exhibited challenges in bias and uncertainty calibration, reflected in low rejection accuracies and high entropy scores. High entropy scores in Janus models (Janus 7B: 0.8392, Janus 1B: 0.787) underscore their susceptibility to positional bias and unstable reasoning, contrasting with the low entropy and robust reasoning of ChatGPT models. The study further demonstrates that model size is not the sole determinant of performance, as evidenced by Grok 3 underperformance despite its substantial parameter count. By employing multi-image contexts, rejection mechanisms, and entropy-based consistency metrics, this benchmark sets a new standard for evaluating multimodal LLMs, enabling a more robust and reliable assessment of next-generation AI systems.

Large Language Models (LLMs) have demonstrated strong translation abilities through prompting, even without task-specific training. However, their effectiveness in dialectal and low-resource contexts remains underexplored. This study presents the first systematic investigation of LLM-based machine translation (MT) for Sylheti, a dialect of Bangla that is itself low-resource. We evaluate five advanced LLMs (GPT-4.1, GPT-4.1, LLaMA 4, Grok 3, and DeepSeek V3.2) across both translation directions (Bangla Leftrightarrow Sylheti), and find that these models struggle with dialect-specific vocabulary. To address this, we introduce Sylheti-CAP (Context-Aware Prompting), a three-step framework that embeds a linguistic rulebook, a dictionary (2{,}260 core vocabulary items and idioms), and an authenticity check directly into prompts. Extensive experiments show that Sylheti-CAP consistently improves translation quality across models and prompting strategies. Both automatic metrics and human evaluations confirm its effectiveness, while qualitative analysis reveals notable reductions in hallucinations, ambiguities, and awkward phrasing, establishing Sylheti-CAP as a scalable solution for dialectal and low-resource MT. Dataset link: https://github.com/TabiaTanzin/LLMs-for-Low-Resource-Dialect-Translation-Using-Context-Aware-Prompting-A-Case-Study-on-Sylheti.git{https://github.com/TabiaTanzin/LLMs-for-Low-Resource-Dialect-Translation-Using-Context-Aware-Prompting-A-Case-Study-on-Sylheti.git}

This paper introduces a novel Deep Researcher architecture designed to generate detailed research reports on complex PhD level topics by addressing the inherent limitations of the Parallel Scaling paradigm. Our system utilizes two key innovations: Sequential Research Plan Refinement via Reflection and a Candidates Crossover algorithm. The sequential refinement process is demonstrated as an efficient method that allows the agent to maintain a centralized Global Research Context, enabling it to look back at current progress, reason about the research plan, and intelligently make changes at runtime. This dynamic adaptation contrasts with parallel approaches, which often suffer from siloed knowledge. The Candidates Crossover algorithm further enhances search efficiency by deploying multiple LLM candidates with varied parameters to explore a larger search space, with their findings synthesized to curate a comprehensive final research response. The process concludes with One Shot Report Generation, ensuring the final document is informed by a unified narrative and high fact density. Powered by the Gemini 2.5 Pro model, our Deep Researcher was evaluated on the DeepResearch Bench, a globally recognized benchmark of 100 doctoral level research tasks. Our architecture achieved an overall score of 46.21, demonstrating superior performance by surpassing leading deep research agents such as Claude Researcher, Nvidia AIQ Research Assistant, Perplexity Research, Kimi Researcher and Grok Deeper Search present on the DeepResearch Bench actively running leaderboard. This performance marginally exceeds our previous work, Static DRA, and reinforces the finding that sequential scaling consistently outperforms the parallel self consistency paradigm.

Can machines truly think, reason and act in domains like humans? This enduring question continues to shape the pursuit of Artificial General Intelligence (AGI). Despite the growing capabilities of models such as GPT-4.5, DeepSeek, Claude 3.5 Sonnet, Phi-4, and Grok 3, which exhibit multimodal fluency and partial reasoning, these systems remain fundamentally limited by their reliance on token-level prediction and lack of grounded agency. This paper offers a cross-disciplinary synthesis of AGI development, spanning artificial intelligence, cognitive neuroscience, psychology, generative models, and agent-based systems. We analyze the architectural and cognitive foundations of general intelligence, highlighting the role of modular reasoning, persistent memory, and multi-agent coordination. In particular, we emphasize the rise of Agentic RAG frameworks that combine retrieval, planning, and dynamic tool use to enable more adaptive behavior. We discuss generalization strategies, including information compression, test-time adaptation, and training-free methods, as critical pathways toward flexible, ___domain-agnostic intelligence. Vision-Language Models (VLMs) are reexamined not just as perception modules but as evolving interfaces for embodied understanding and collaborative task completion. We also argue that true intelligence arises not from scale alone but from the integration of memory and reasoning: an orchestration of modular, interactive, and self-improving components where compression enables adaptive behavior. Drawing on advances in neurosymbolic systems, reinforcement learning, and cognitive scaffolding, we explore how recent architectures begin to bridge the gap between statistical learning and goal-directed cognition. Finally, we identify key scientific, technical, and ethical challenges on the path to AGI.

Large language models (LLMs) are increasingly deployed as autonomous decision agents in settings with asymmetric error costs: hiring (missed talent vs wasted interviews), medical triage (missed emergencies vs unnecessary escalation), and fraud detection (approved fraud vs declined legitimate payments). The dominant design queries a single LLM for a posterior over states, thresholds "confidence," and acts; we prove this is inadequate for sequential decisions with costs. We propose a Bayesian, cost-aware multi-LLM orchestration framework that treats LLMs as approximate likelihood models rather than classifiers. For each candidate state, we elicit likelihoods via contrastive prompting, aggregate across diverse models with robust statistics, and update beliefs with Bayes rule under explicit priors as new evidence arrives. This enables coherent belief updating, expected-cost action selection, principled information gathering via value of information, and fairness gains via ensemble bias mitigation. In resume screening with costs of 40000 USD per missed hire, 2500 USD per interview, and 150 USD per phone screen, experiments on 1000 resumes using five LLMs (GPT-4o, Claude 4.5 Sonnet, Gemini Pro, Grok, DeepSeek) reduce total cost by 294000 USD (34 percent) versus the best single-LLM baseline and improve demographic parity by 45 percent (max group gap 22 to 5 percentage points). Ablations attribute 51 percent of savings to multi-LLM aggregation, 43 percent to sequential updating, and 20 percent to disagreement-triggered information gathering, consistent with the theoretical benefits of correct probabilistic foundations.

The integration of large language models (LLMs) on low-power edge devices such as Raspberry Pi, known as edge language models (ELMs), has introduced opportunities for more personalized, secure, and low-latency language intelligence that is accessible to all. However, the resource constraints inherent in edge devices and the lack of robust ethical safeguards in language models raise significant concerns about fairness, accountability, and transparency in model output generation. This paper conducts a comparative analysis of text-based bias across language model deployments on edge, cloud, and desktop environments, aiming to evaluate how deployment settings influence model fairness. Specifically, we examined an optimized Llama-2 model running on a Raspberry Pi 4; GPT 4o-mini, Gemini-1.5-flash, and Grok-beta models running on cloud servers; and Gemma2 and Mistral models running on a MacOS desktop machine. Our results demonstrate that Llama-2 running on Raspberry Pi 4 is 43.23% and 21.89% more prone to showing bias over time compared to models running on the desktop and cloud-based environments. We also propose the implementation of a feedback loop, a mechanism that iteratively adjusts model behavior based on previous outputs, where predefined constraint weights are applied layer-by-layer during inference, allowing the model to correct bias patterns, resulting in 79.28% reduction in model bias.

Text-to-Image (TTI) models are powerful creative tools but risk amplifying harmful social biases. We frame representational societal bias assessment as an image curation and evaluation task and introduce a pilot benchmark of occupational portrayals spanning five socially salient roles (CEO, Nurse, Software Engineer, Teacher, Athlete). Using five state-of-the-art models: closed-source (DALLE 3, Gemini Imagen 4.0) and open-source (FLUX.1-dev, Stable Diffusion XL Turbo, Grok-2 Image), we compare neutral baseline prompts against fairness-aware controlled prompts designed to encourage demographic diversity. All outputs are annotated for gender (male, female) and race (Asian, Black, White), enabling structured distributional analysis. Results show that prompting can substantially shift demographic representations, but with highly model-specific effects: some systems diversify effectively, others overcorrect into unrealistic uniformity, and some show little responsiveness. These findings highlight both the promise and the limitations of prompting as a fairness intervention, underscoring the need for complementary model-level strategies. We release all code and data for transparency and reproducibility https://github.com/maximus-powers/img-gen-bias-analysis.

Large language models (LLMs) can perform reasoning computations both internally within their latent space and externally by generating explicit token sequences like chains of thought. Significant progress in enhancing reasoning abilities has been made by scaling test-time compute. However, understanding and quantifying model-internal reasoning abilities - the inferential "leaps" models make between individual token predictions - remains crucial. This study introduces a benchmark (n = 4,000 items) designed to quantify model-internal reasoning in different domains. We achieve this by having LLMs indicate the correct solution to reasoning problems not through descriptive text, but by selecting a specific language of their initial response token that is different from English, the benchmark language. This not only requires models to reason beyond their context window, but also to overrise their default tendency to respond in the same language as the prompt, thereby posing an additional cognitive strain. We evaluate a set of 18 LLMs, showing significant performance variations, with GPT-4.5 achieving the highest accuracy (74.7%), outperforming models like Grok-2 (67.2%), and Llama 3.1 405B (65.6%). Control experiments and difficulty scaling analyses suggest that while LLMs engage in internal reasoning, we cannot rule out heuristic exploitations under certain conditions, marking an area for future investigation. Our experiments demonstrate that LLMs can "think" via latent-space computations, revealing model-internal inference strategies that need further understanding, especially regarding safety-related concerns such as covert planning, goal-seeking, or deception emerging without explicit token traces.

Alignment faking in large language models presented a demonstration of Claude 3 Opus and Claude 3.5 Sonnet selectively complying with a helpful-only training objective to prevent modification of their behavior outside of training. We expand this analysis to 25 models and find that only 5 (Claude 3 Opus, Claude 3.5 Sonnet, Llama 3 405B, Grok 3, Gemini 2.0 Flash) comply with harmful queries more when they infer they are in training than when they infer they are in deployment. First, we study the motivations of these 5 models. Results from perturbing details of the scenario suggest that only Claude 3 Opus's compliance gap is primarily and consistently motivated by trying to keep its goals. Second, we investigate why many chat models don't fake alignment. Our results suggest this is not entirely due to a lack of capabilities: many base models fake alignment some of the time, and post-training eliminates alignment-faking for some models and amplifies it for others. We investigate 5 hypotheses for how post-training may suppress alignment faking and find that variations in refusal behavior may account for a significant portion of differences in alignment faking.

Recent advances in multi-modal Large Language Models (M-LLMs) have demonstrated a powerful ability to synthesize implicit information from disparate sources, including images and text. These resourceful data from social media also introduce a significant and underexplored privacy risk: the inference of sensitive personal attributes from seemingly daily media content. However, the lack of benchmarks and comprehensive evaluations of state-of-the-art M-LLM capabilities hinders the research of private attribute profiling on social media. Accordingly, we propose (1) PRISM, the first multi-modal, multi-dimensional and fine-grained synthesized dataset incorporating a comprehensive privacy landscape and dynamic user history; (2) an Efficient evaluation framework that measures the cross-modal privacy inference capabilities of advanced M-LLM. Specifically, PRISM is a large-scale synthetic benchmark designed to evaluate cross-modal privacy risks. Its key feature is 12 sensitive attribute labels across a diverse set of multi-modal profiles, which enables targeted privacy analysis. These profiles are generated via a sophisticated LLM agentic workflow, governed by a prior distribution to ensure they realistically mimic social media users. Additionally, we propose a Multi-Agent Inference Framework that leverages a pipeline of specialized LLMs to enhance evaluation capabilities. We evaluate the inference capabilities of six leading M-LLMs (Qwen, Gemini, GPT-4o, GLM, Doubao, and Grok) on PRISM. The comparison with human performance reveals that these MLLMs significantly outperform in accuracy and efficiency, highlighting the threat of potential privacy risks and the urgent need for robust defenses.

The Internet of Things has expanded rapidly, transforming communication and operations across industries but also increasing the attack surface and security breaches. Artificial Intelligence plays a key role in securing IoT, enabling attack detection, attack behavior analysis, and mitigation suggestion. Despite advancements, evaluations remain purely qualitative, and the lack of a standardized, objective benchmark for quantitatively measuring AI-based attack analysis and mitigation hinders consistent assessment of model effectiveness. In this work, we propose a hybrid framework combining Machine Learning (ML) for multi-class attack detection with Large Language Models (LLMs) for attack behavior analysis and mitigation suggestion. After benchmarking several ML and Deep Learning (DL) classifiers on the Edge-IIoTset and CICIoT2023 datasets, we applied structured role-play prompt engineering with Retrieval-Augmented Generation (RAG) to guide ChatGPT-o3 and DeepSeek-R1 in producing detailed, context-aware responses. We introduce novel evaluation metrics for quantitative assessment to guide us and an ensemble of judge LLMs, namely ChatGPT-4o, DeepSeek-V3, Mixtral 8x7B Instruct, Gemini 2.5 Flash, Meta Llama 4, TII Falcon H1 34B Instruct, xAI Grok 3, and Claude 4 Sonnet, to independently evaluate the responses. Results show that Random Forest has the best detection model, and ChatGPT-o3 outperformed DeepSeek-R1 in attack analysis and mitigation.