Daily Papers

Evaluating the occurrence and triggers of large language model (LLM) behaviors - such as sycophancy, self-preference, or over-confidence - is critical for predicting real-world model deployment risks. However, existing situated behavioral evaluations typically ignore cultural context, limiting their generalizability across an increasingly global user base. To address this gap, we propose CuBEs - Culturally-situated Behavior Evaluations that probe for response patterns across diverse user cultures. We first extend an automated testing pipeline to inject cultural context into behavioral test scenarios and subsequent evaluation. We assess the cultural adaptability of this pipeline by building a human-labeled dataset that captures nuanced dimensions of behavior understanding across 12 distinct cultures. Our dataset reveals significant cross-cultural variations that one-size-fits all judgments fail to capture. Through evaluating 13 open- and closed-source LLMs, we find that introducing cultural situatedness in the evaluation scenario creates significant variation in the presence of a behavior. For example, while our baseline experiments testing for political bias capture localized Western political dimensions like the American conservative-progressive divide, non-Western culturally situated evaluations surface entirely different axes of bias such as religious and colonial political issues. Our findings demonstrate that standard, culturally-agnostic evaluations fail to capture these shifts, highlighting the necessity of culturally situated behavioral testing for global deployments.

The task of "unlearning" certain concepts in large language models (LLMs) has attracted immense attention recently, due to its importance for mitigating undesirable model behaviours, such as the generation of harmful, private, or incorrect information. Current protocols to evaluate unlearning methods largely rely on behavioral tests, without monitoring the presence of unlearned knowledge within the model's parameters. This residual knowledge can be adversarially exploited to recover the erased information post-unlearning. We argue that unlearning should also be evaluated internally, by considering changes in the parametric knowledge traces of the unlearned concepts. To this end, we propose a general methodology for eliciting directions in the parameter space (termed "concept vectors") that encode concrete concepts, and construct ConceptVectors, a benchmark dataset containing hundreds of common concepts and their parametric knowledge traces within two open-source LLMs. Evaluation on ConceptVectors shows that existing unlearning methods minimally impact concept vectors, while directly ablating these vectors demonstrably removes the associated knowledge from the LLMs and significantly reduces their susceptibility to adversarial manipulation. Our results highlight limitations in behavioral-based unlearning evaluations and call for future work to include parametric-based evaluations. To support this, we release our code and benchmark at https://github.com/yihuaihong/ConceptVectors.

Large language models can memorize private or harmful information, motivating machine unlearning methods that remove targeted knowledge while preserving other capabilities. However, existing evaluations rely primarily on behavioral benchmarks, which assess whether a model appears to forget but provide limited insight into how unlearning changes the model or affects related knowledge. We introduce TRIAGE (Tripartite Representation-internal Introspection for Adjacency Gap Evaluation), a benchmark-agnostic evaluation framework for characterizing these changes. TRIAGE uses diagonal approximations of the Fisher information and Hessian to measure changes in parameter sensitivity and local curvature, and utilizes a Forget / Adjacent-Retain / Generic-Retain partition to quantify an adjacency gap in semantically related knowledge. Based on the magnitude and distribution of these changes, TRIAGE further classifies each algorithm's update as no-op, partially localized, collateral dominant, or globally destructive. Across 12 unlearning methods, four language models, and the WMDP, TOFU, and MUSE benchmarks, we find that methods with similar behavioral forgetting can produce substantially different internal changes and patterns of collateral damage. These signatures also vary across models and benchmarks, indicating that the effects of unlearning are not determined solely by the unlearning algorithm. TRIAGE can be applied alongside existing unlearning benchmarks to complement behavioral evaluation with a model-internal view of how unlearning reshapes the model's parameter space and affects retained knowledge.

Compositional generalization-the ability to interpret novel combinations of familiar components-remains a persistent challenge for neural networks. Behavioral evaluations reveal when models fail but offer limited insight into why failures arise at the representational level. We introduce Homomorphism Error (HE), a structural metric that quantifies deviations from approximate homomorphisms between the expression algebra and a model's hidden-state space. We instantiate HE for two compositional operators in SCAN-style tasks: modifier HE for unary composition and sequence HE for binary composition, measured by learning representation-level operators that predict composed representations from their parts. Across controlled experiments with small decoder-only Transformers, HE predicts out-of-distribution (OOD) compositional generalization under noise injection, achieving R^2 = 0.73 correlation between modifier HE and OOD accuracy. Ablations show that model depth has minimal effect on either HE or OOD accuracy, training data coverage exhibits threshold effects (insufficient coverage sharply increases HE and degrades OOD performance), and randomly inserted noise tokens systematically increase HE. Finally, we test if HE-regularized training improves OOD accuracy. Experiment shows that explicitly enforcing low modifier HE during training significantly reduces modifier HE (p = 1.1x10-4) and sequence HE (p = 0.001) and yields a statistically significant improvement in OOD accuracy (p = 0.023). Together, these results indicate the potential of HE to be both a diagnostic and an actionable training signal for improving compositional generalization. Code to reproduce our experiments is open-sourced.

Existing benchmarks for LLM-based offensive security agents use isolated, single-target setups with a known vulnerable service and fixed objective. They measure exploitation effectively, but miss how real Capture-the-Flag (CTF) participants triage unknown surfaces, prioritize targets, and allocate effort under uncertainty. Current evaluations therefore fail to assess strategic reasoning beyond exploitation alone. To address this, we introduce CTFExplorer, a benchmark suite that shifts offensive security evaluation toward a multi-target setting, which tests how agents explore, prioritize, and chain attacks. CTFExplorer deploys 40 web-based vulnerable services within a single environment, where agents must autonomously discover, distinguish, and exploit targets without predefined guidance. We also present a reactive multi-agent setup as a reference agent framework and develop an agent-agnostic evaluation framework that records structured reasoning traces for fine-grained assessment. This enables behavioral evaluation beyond binary flag capture, such as how agents manage target selection, handle failed hypotheses, coordinate across multiple stages, and extract security intelligence.

Generative CAD models are expected to remain behaviorally correct after parameter edits, so increasing the number of edit checks is often treated as a direct route to more reliable evaluation. Under a fixed budget, however, auditing each program more thoroughly reduces the number of tasks and independent generations that can be evaluated, which can ultimately make model-level estimates less accurate. We study this phenomenon and the conditions under which it arises. We decompose behavioral evaluation into three evidence levels: task templates, stochastic generations, and within-program edits. We define an average failure risk that is invariant to audit depth, and combine three-level variance with measured execution costs to analyze the tradeoff between deeper edit auditing and broader independent coverage. Experiments across two CAD environments and five generation systems show that the value of deeper auditing depends on where evaluation uncertainty originates. When template heterogeneity or generation stochasticity dominates, additional edit checks can increase total estimation error; when within-program state variation is large and generation is expensive, deeper auditing is more valuable. Variance and cost estimates from calibration predict the direction of this change and provide a diagnostic basis for allocating evidence on held-out tasks. These results show that the thoroughness of program inspection can diverge from the reliability of model evaluation, and they help determine whether the next unit of budget should be spent on a new task, a new generation, or additional edit checks.

We address whether neural models for Natural Language Inference (NLI) can learn the compositional interactions between lexical entailment and negation, using four methods: the behavioral evaluation methods of (1) challenge test sets and (2) systematic generalization tasks, and the structural evaluation methods of (3) probes and (4) interventions. To facilitate this holistic evaluation, we present Monotonicity NLI (MoNLI), a new naturalistic dataset focused on lexical entailment and negation. In our behavioral evaluations, we find that models trained on general-purpose NLI datasets fail systematically on MoNLI examples containing negation, but that MoNLI fine-tuning addresses this failure. In our structural evaluations, we look for evidence that our top-performing BERT-based model has learned to implement the monotonicity algorithm behind MoNLI. Probes yield evidence consistent with this conclusion, and our intervention experiments bolster this, showing that the causal dynamics of the model mirror the causal dynamics of this algorithm on subsets of MoNLI. This suggests that the BERT model at least partially embeds a theory of lexical entailment and negation at an algorithmic level.

LLMs memorize sensitive training data, including personally identifiable information (PII), creating a pressing need for reliable post hoc removal methods. Unlearning has emerged as a promising solution, with state-of-the-art (SOTA) methods often following a localize-first, unlearn-second paradigm that targets specific model parameters. However, existing benchmarks evaluate unlearning solely at the output level, leaving open the question of whether unlearning truly erases knowledge from a model's parameters or merely obfuscates it, a concern reinforced by the success of resurfacing attacks. To bridge this gap, we introduce LACUNA: the first unlearning testbed with ground-truth parameter-level localization. LACUNA injects PII of synthetic individuals into predefined parameters of 1B and 7B OLMo-based models via masked continual pretraining, enabling direct evaluation of whether unlearning targets the weights responsible for knowledge storage. We use LACUNA to benchmark current SOTA unlearning methods and find that, despite strong output-level performance, existing methods are highly imprecise and susceptible to resurfacing attacks. We further show that when localization is successful, even a simple gradient-based unlearning method achieves strong erasure and robustness to resurfacing attacks, highlighting the importance of precise unlearning. We release LACUNA to complement behavioral evaluations and drive further advances in robust, localization-based unlearning.

Emergent misalignment (EM) occurs when narrow finetuning causes a model to behave dangerously outside the finetuning task. Standard training signals can miss this shift, making reliable detection costly if it depends on repeated behavioral evaluation. We ask whether emergent misalignment can instead be detected from internal representations during finetuning. Using seven alignment-relevant traits encoded as linear directions in activation space, we track representational drift across training checkpoints in four open-source 7-9B LLMs. EM-relevant drift concentrates on a low-dimensional axis that explains 65.5% of the variance, revealing a geometric signature in the studied regime. A low-overhead monitor built on this drift profile detects dangerous checkpoints with 2.2% false negative rate, 2.9% false positive rate, and 0.990 AUROC on held-out perturbation types, outperforming unsupervised PCA and SAE baselines. Stress tests on two 14B models, longer finetuning runs, and misaligned starting points identify key deployment boundaries. These results position trait-space monitoring as a practical complement to behavioral evaluation for EM detection during LoRA-based finetuning, while showing that deployment across substantially different regimes may require recalibration.

Post-training quantization is widely used to deploy large language models in resource-constrained settings, yet its evaluation relies almost exclusively on accuracy and perplexity. We show that these metrics fail to capture behavioral changes induced by quantization. We introduce correctness agreement, a decision-level metric that measures overlap in correct predictions between a base model and its quantized variants, independent of absolute accuracy. Across multiple models and quantization schemes from 8-bit to 2-bit, we find that behavioral divergence emerges under moderate quantization even when task performance appears preserved. To explain this effect, we analyze quantization as a structural operator on attention weights and quantify layer-wise distortions using statistical and distributional measures. Our results reveal non-linear breakpoints at low bit-widths and show that query and key projections are consistently more sensitive than value and output projections. These findings expose an illusion of equivalence between base and quantized models and motivate behavioral evaluation beyond conventional performance metrics.

Large language models are increasingly used in clinical settings. However, research into the reliability and performance of these models has focused almost entirely on the language substrate, scoring what the model says. Mechanistic interpretability has found that the latent space carries a higher fidelity of representation than the text: internal representations not only encode substantially more than the output verbalizes, but the stated reasoning also systematically omits features that causally drive the answer. An evaluation of model behavior in terms of mechanistic interpretability has not been explored in clinical decision support. In this work, we extend behavioral evaluation into the latent space and ask whether clinical concepts exist as locatable, causally used representations inside open-weight LLMs. We find dedicated clinical concept centers in the latent space of all eleven open models we test. These concept centers are interpretable, firing only on their aligned clinical narratives, and meaningfully and causally drive model behavior in both constrained and open-ended settings. They are not just analytical representations, but circuits that can be utilized in clinical practice, and we explore their use from the perspective of both evaluation and performance. From the evaluation standpoint, models stay internally coherent and keep using the relevant concept centers even under adversarial role-based priming, while aligned priming improves downstream clinical performance. From a performance perspective, we simulate realistic deployment settings and find that steering models along these centers leads to meaningful downstream improvements. Finally, we conduct a blinded clinician validation and find the activation and usage of these concept centers predicts clinicians preferences.

Training reliable tool-augmented agents remains a significant challenge, largely due to the difficulty of credit assignment in multi-step reasoning. While process-level reward models offer a promising direction, existing LLM-based judges often produce noisy and inconsistent signals because they lack fine-grained, task-specific rubrics to distinguish high-level planning from low-level execution. In this work, we introduce SCRIBE (Skill-Conditioned Reward with Intermediate Behavioral Evaluation), a reinforcement learning framework that intervenes at a novel mid-level abstraction. SCRIBE grounds reward modeling in a curated library of skill prototypes, transforming open-ended LLM evaluation into a constrained verification problem. By routing each subgoal to a corresponding prototype, the reward model is equipped with precise, structured rubrics that substantially reduce reward variance. Experimental results show that SCRIBE achieves state-of-the-art performance across a range of reasoning and tool-use benchmarks. In particular, it improves the AIME25 accuracy of a Qwen3-4B model from 43.3% to 63.3%, and significantly increases success rates in complex multi-turn tool interactions. Further analysis of training dynamics reveals a co-evolution across abstraction levels, where mastery of mid-level skills consistently precedes the emergence of effective high-level planning behaviors. Finally, we demonstrate that SCRIBE is additive to low-level tool optimizations, providing a scalable and complementary pathway toward more autonomous and reliable tool-using agents.

A model that lies while knowing the truth is the central case ELK cannot handle with behavioral evaluation alone. We ask whether such deception leaves an internal signature distinguishing it from honest error. Our key move is a control for wrongness: we contrast a sleeper agent (knows the truth, lies on trigger) against a naive liar (fine-tuned to emit the same wrong answers with no honest training). Both produce identical wrong outputs; any difference is about knowledge conflict, not incorrectness. We find deceptive forward passes carry a conflict signature - 2.1-2.3x higher residual rank than naive-liar passes on the same wrong answer - strong enough to identify which of two responses is the lie with 100% accuracy and no labels, across GPT-2 small/medium (three seeds) and three instruct models. Across Qwen2.5-1.5B/7B and Phi-3-mini, instructed deception raises residual rank on every tested fact (18/18, 40/40, 34/34); on Phi-3, lies separate perfectly from both honest answers and hallucinations (AUC 1.0, Wilcoxon p~6e-11). The signature survives strategic self-constructed deception (model invents its own lie, AUC 1.0), active concealment attempts (AUC 1.0), and length-controlled replication (20/20, AUC 1.0, p~1e-6). Using basis-free relative representations, a probe trained on one model family detects deception in two other families zero-shot (mean AUC 0.933), surviving simultaneous architecture and format change (AUC 0.821), and transfers across five languages (AUC 1.000, length-controlled). The signature is read-only: detectable but not injectable (0/8 both directions). Honest limitations and six negative experiments are documented in full.

Large language models typically undergo post-training to align them with safety policies but there exist many sophisticated jailbreaks that sidestep established safeguards. For instance, prior work by Andriushchenko et al. (2025) has found that changing the grammatical tense from present to past can be enough to elicit harmful responses. In this work, we uncover a more general failure of non-imperative syntactic forms. We demonstrate that this syntactic vulnerability exists in 16 models up to 70B parameters, using behavioral evaluation. To investigate the root cause, we apply causal mediation analysis, finding that refusal is partially conditioned on upstream syntactic features. By steering these purely syntactic features we are able to trigger and suppress refusal. Finally, we trace this ill-conditioning to linguistically biased post-training data of open-source models and show that increasing syntactic diversity can mitigate the issue. Our findings suggest that current alignment approaches introduce confounders that prevent a pure semantic grounding of the refusal decision.

Biosecurity evaluations of language models typically ask whether models produce hazardous output. This paper asks a complementary question: when a model refuses, is that refusal structurally sound, or does it disappear under modest changes to prompt framing, formatting, or output length? Across five architectures, no model cleanly discriminated benign from hazard. Gemma 2 2B-IT never genuinely refused across 75 prompts, hedging on every hazard-adjacent query. Gemma 4 E2B-IT refused 65/75 prompts with chat-template formatting and 0/75 without it. Both Gemma models collapsed to 0% under an 80-token cap. Qwen 2.5 1.5B and Phi-3-mini over-refused, flagging 83-87% of benign biology as hazardous. Llama 3.2 1B showed the only meaningful tier gradient (61-point spread). To probe what drives such over-refusal, we tested a panel of Schedule I but biologically non-toxic compounds (notably psilocybin cultivation, with FDA Breakthrough Therapy status). Some models refused these at rates exceeding genuinely hazardous biology, suggesting refusal tracks legality and cultural salience over CBRN hazard. To measure the internal side, we introduce a divergence score D comparing a model's surface response label to its internal sparse autoencoder (SAE) feature activations. Full D was computed on Gemma 2 2B-IT (Gemma Scope 1) and Gemma 4 E2B-IT (author-trained bio SAE). Two fine-tuned Gemma 2 domain SAEs were released. On Gemma 4, comply and refuse responses separated by a 0.647-point gap with zero overlap (n=75), though this is preliminary, with a narrow catalog, within-sample calibration, and Gemma-family-only SAE coverage. Built over one hackathon weekend on consumer hardware (GTX 1650 Ti Max-Q, plus Colab T4 for SAE training), this preliminary evidence suggests activation-level auditing may surface failure modes invisible to behavioral evaluation, with substantial variation across architectures.

Large Language Model (LLM) safety has often been evaluated at the behavior level, which provides limited evidence of internal robustness, as these evaluations target outputs rather than representation-level vulnerability under intervention. We formalize this discrepancy as the audit gap: the difference between behavioral safety and robustness under intervention. To study this gap, we construct dissociated models that preserve safe outward behavior while remaining vulnerable in the latent space. We introduce an intervention-based evaluation framework to test model robustness through soft interventions in parameter and latent spaces, including harmful fine-tuning and layer-wise latent perturbations. To formalize the evaluation, we propose the Latent Vulnerability Score (LVS) to measure how easily harmful behavior can be elicited by bounded latent perturbations. Using this evaluation framework, we show that behavioral safety metrics are insufficient measures of representation-level robustness across multiple safely and unsafely aligned state-of-the-art models. Notably, dissociated models show substantially elevated LVSs despite comparable refusal behavior under harmful intervention, with intermediate representations being the most sensitive to intervention. Our results suggest that behavioral safety evaluation alone provides an incomplete picture of model robustness, motivating representation-aware audits of latent vulnerability and observable behavior.

The rapid proliferation of LLM-based autonomous agents in real operating system environments introduces a new category of safety risk beyond content safety: behavior jailbreak, where an adversary induces an agent to execute dangerous OS-level operations with irreversible consequences. Existing benchmarks either evaluate safety at the semantic layer alone, missing physical-layer harms, or fail to isolate test cases, letting earlier runs contaminate later ones. We present LITMUS (LLM-agents In-OS Testing for Measuring Unsafe Subversion), a benchmark addressing both gaps via a semantic-physical dual verification mechanism and OS-level state rollback. LITMUS comprises 819 high-risk test cases organized into one harmful seed subset and six attack-extended subsets covering three adversarial paradigms (jailbreak speaking, skill injection, and entity wrapping), plus a fully automated multi-agent evaluation framework judging behavior at both conversational and OS-level physical layers. Evaluation across frontier agents reveals three findings: (1) current agents lack effective safety awareness, with strong models (e.g., Claude Sonnet 4.6) still executing 40.64% of high-risk operations; (2) agents exhibit pervasive Execution Hallucination (EH), verbally refusing a request while the dangerous operation has already completed at the system level, invisible to every prior semantic-only framework; and (3) skill injection and entity wrapping attacks achieve high success rates, exposing pronounced agent vulnerabilities. LITMUS provides the first standardized platform for reproducible, physically grounded behavioral safety evaluation of LLM agents in real OS environments.

As reasoning capacity and deployment scope grow in tandem, large language models (LLMs) gain the capacity to engage in behaviors that serve their own objectives, a class of risks we term Emergent Strategic Reasoning Risks (ESRRs). These include, but are not limited to, deception (intentionally misleading users or evaluators), evaluation gaming (strategically manipulating performance during safety testing), and reward hacking (exploiting misspecified objectives). Systematically understanding and benchmarking these risks remains an open challenge. To address this gap, we introduce ESRRSim, a taxonomy-driven agentic framework for automated behavioral risk evaluation. We construct an extensible risk taxonomy of 7 categories, which is decomposed into 20 subcategories. ESRRSim generates evaluation scenarios designed to elicit faithful reasoning, paired with dual rubrics assessing both model responses and reasoning traces, in a judge-agnostic and scalable architecture. Evaluation across 11 reasoning LLMs reveals substantial variation in risk profiles (detection rates ranging 14.45%-72.72%), with dramatic generational improvements suggesting models may increasingly recognize and adapt to evaluation contexts.

Large Language Models (LLMs) are increasingly deployed in socially sensitive domains, yet their unpredictable behaviors, ranging from misaligned intent to inconsistent personality, pose significant risks. We introduce SteerEval, a hierarchical benchmark for evaluating LLM controllability across three domains: language features, sentiment, and personality. Each domain is structured into three specification levels: L1 (what to express), L2 (how to express), and L3 (how to instantiate), connecting high-level behavioral intent to concrete textual output. Using SteerEval, we systematically evaluate contemporary steering methods, revealing that control often degrades at finer-grained levels. Our benchmark offers a principled and interpretable framework for safe and controllable LLM behavior, serving as a foundation for future research.

Using a large language model to play the user is now standard in scalable evaluation. It has a repeatedly diagnosed failure: the simulated user is excessively cooperative, so a system under test can score by the sheer number of questions it asks rather than by making the user willing to speak. We answer with a disclosure gate conditioning information release on the companion agent's behaviour: its state is a ladder of five ordered gates, merged onto three observable depth layers. We specify, ablate, and audit it, and train a user simulator against that specification. Gating behaviour is learned from the training corpus's synthetic branch, while the real branch supplies how people speak and react; after training, the simulator need not be told at runtime which gate each item sits behind. The gate is a load-bearing component of the environment: on the English corpus of a published companion-agent benchmark (CompanionBench), once training no longer states per example which gate each item sits behind, the largest rank displacement across 12 systems under test exceeds the noise band set by re-running that environment under a new seed, while per-system scores show no detectable change. We state two acceptance criteria: a ranking must be order-preserving, and absolute scores must be scale-stable. Of the candidates we examine, only one passes both -- the simulator we release -- and its leaderboard correlates at 0.993 with the benchmark's original simulator. By contrast, prompting a frontier model as the simulator barely moves the ranking while shifting every score upward -- a shift invisible to anyone checking the ranking alone. The environment we specify is the one that benchmark already used. That publication describes the mechanism in about four hundred words, and we supply what it lacked: specification, ablations, human studies, negative controls, and downstream sensitivity analysis.

Standard unlearning evaluations measure behavioral suppression in full precision, immediately after training, despite every deployed language model being quantized first. Recent work has shown that 4-bit post-training quantization can reverse machine unlearning; we show this is not a tuning artefact but a systematic dual failure: gradient-based methods that achieve meaningful forgetting lose it under compression, while methods that survive quantization barely change the model. Both failures trace to the same root cause: across all baselines, per-parameter updates lie 47-828x below the NF4 quantization bin width; updates diffused across billions of parameters cannot clear quantization bin boundaries, a consequence we formalize as a sparsity-permanence tradeoff. We present MANSU (Mechanistic-Aligned Null-Space Unlearning), which resolves both modes by combining causal circuit attribution to isolate the minimal forget-set subgraph, circuit-restricted null-space projection with a diagonal-Fisher retain bound, and a per-parameter magnitude floor guaranteeing quantization survival by construction. We additionally introduce Circuit Attribution Divergence (CAD), a mechanistic verification metric distinguishing structural erasure from behavioral suppression, a distinction existing metrics cannot make. Across multiple model families and hazard benchmarks, MANSU is the first method to jointly satisfy all four properties with margin on each (meaningful forgetting, retain preservation, non-positive PTQ gap, and structural erasure), while gradient-based baselines recover up to +0.05 accuracy under compression.

Evaluating retail strategies before deployment is difficult, as outcomes are determined across multiple stages, from seller-side persuasion through buyer-seller interaction to purchase decisions. However, existing retail simulators capture only partial aspects of this process and do not model cross-stage dependencies, making it difficult to assess how early decisions affect downstream outcomes. We present RetailSim, an end-to-end retail simulation framework that models this pipeline in a unified environment, explicitly designed for simulation fidelity through diverse product spaces, persona-driven agents, and multi-turn interactions. We evaluate RetailSim with a dual protocol comprising human evaluation of behavioral fidelity and meta-evaluation against real-world economic regularities, showing that it successfully reproduces key patterns such as demographic purchasing behavior, the price-demand relationship, and heterogeneous price elasticity. We further demonstrate its practical utility via decision-oriented use cases, including persona inference, seller-buyer interaction analysis, and sales strategy evaluation, showing RetailSim's potential as a controlled testbed for exploring retail strategies.

Large language models (LLMs) are increasingly deployed in settings that require nuanced ethical reasoning, yet existing bias evaluations treat model outputs as simply "biased" or "unbiased." This binary framing misses the gradual, context-sensitive way bias actually emerges. We address this gap in two stages: behavioral profiling and mechanistic validation. In the behavioral stage, we introduce the Moral Sensitivity Index (MSI), a metric that quantifies the probability of biased output across a graduated, seven-tier stress test ranging from abstract numerical problems to scenarios rooted in historical and socioeconomic injustice. Evaluating four leading models (Claude 3.5, Qwen 3.5, Llama 3, and Gemini 1.5), we identify distinct behavioral signatures shaped by alignment design: for instance, Gemini 1.5 reaches 72.7% MSI by Tier 5 under socioeconomic framing, while Claude exhibits sharp suppression consistent with identity-based safety training. We then verify these behavioral patterns mechanistically. We select criminal-bias scenarios, which produced the highest MSI scores across models, as probes and apply logit lens, attention analysis, activation patching, and semantic probing to a controlled set of six models spanning three capability tiers: small language models (SLMs), instruction-tuned base models, and reasoning-distilled variants. Circuit-level analysis reveals a U-curve of bias: SLMs exhibit strong criminal bias; scaling to instruction-tuned models eliminates it; reasoning distillation reintroduces bias to SLM-like levels despite identical parameter counts, suggesting distillation compresses reasoning traces in ways that reactivate shallow statistical associations. Critically, the socially loaded cues that drive high MSI scores activate the same bias-driving circuits identified mechanistically, providing cross-stage validation.

Existing full-duplex speech benchmarks cover only subsets of real-time interaction behaviors, often under limited contextual conditions. We introduce DuplexAct-Bench, a bilingual benchmark that systematically covers six complementary behaviors, from interruption and yielding to proactive initiation, active silence, and backchanneling, across Pre-session, In-session, and No-explicit conditions. Across 1,290 English and Chinese streaming trials, we evaluate 12 full-duplex speech systems on both Timing and Content. Results reveal substantial variation across behaviors, conditions, and systems, as well as frequent mismatches between semantic quality and behavioral timing. These findings show that current systems remain far from robustly managing when, whether, and how to participate as real-time interaction unfolds. Project page: https://alitaxky.icu/DuplexAct-Bench/

As large language models are increasingly deployed to simulate diverse human characters, ensuring persona fidelity, defined as the extent to which an agent's behavior consistently reflects the psychological and stylistic characteristics of a target persona, has become a critical requirement. However, existing evaluation paradigms primarily rely on either holistic LLM-based judges, which are prone to "holistic appraisal hallucination'', or static psychometric inventories, which fail to capture the context-dependent fidelity required in dynamic dialogue. To address these limitations, we propose PRISM (Persona Reasoning with Inverse SFL-based Modeling), a psycholinguistically grounded framework that reformulates persona fidelity evaluation as a structured inverse inference task. Inspired by Systemic Functional Linguistics (SFL), PRISM decomposes persona fidelity into three functional dimensions: Task Framing, Interpersonal Stance, and Linguistic Style. It estimates dimension-specific evidence over a persona-conditioned label space and aggregates these signals into an interpretable and auditable evaluation process. Experiments show that PRISM yields more accurate and stable judgements than traditional holistic judging, providing a more reliable framework for persona fidelity evaluation.

Large language model (LLM) agents are increasingly used for complex information-extraction tasks, yet it remains unclear whether agentic components such as reflection and memory lead to observable and controllable improvements over fixed LLM workflows. We study this question through conference-paper dataset extraction, where a system must identify datasets mentioned in scholarly PDFs and produce structured records. We compare a fixed workflow baseline with reflective agent variants and specify an optimized agent condition (S2) that extends the same task with richer PDF tools and dynamic tool selection. Our evaluation emphasizes process-level behavior--including tool execution, retries, reflection, memory use, runtime, and failure recovery--while treating extraction coverage and field completeness as secondary outcome measures. The paper characterizes when agentic mechanisms change system behavior, whether these changes improve task completion, and how the observed failure modes motivate an optimized agent design under the same evaluation harness.

Machine unlearning aims to remove targeted information from large language models, but existing evaluations often measure forgetting only on a designated benchmark distribution. We argue that such behavioral forgetting is an insufficient standard for privacy-oriented unlearning, and that evaluation should instead test whether forgetting persists across the access paths through which the same information may be recovered. We demonstrate this with a controlled case study of cross-lingual querying: across multiple multilingual model families, forgetting varies substantially when semantically equivalent information is queried in different languages and scripts. A single access path is therefore enough to expose a real gap between behavioral forgetting and information inaccessibility, motivating broader evaluation of access-path generalization and adversarial recoverability alongside in-distribution forgetting.

Black-box LLM distillation is usually evaluated as an output-matching problem: a student is considered successful when its responses are semantically similar to, or task-consistent with, those of a teacher. However, output similarity does not imply that the student is behaviorally indistinguishable from the model it imitates. We introduce bounded behavioral indistinguishability, formalized as (ε,q,t,A)-behavioral indistinguishability over an explicit prompt distribution, where ε bounds distinguishing advantage, q bounds oracle queries, t bounds computation, and A denotes the adversary class. We instantiate this notion on Qwen and Llama teacher-student pairs using a controlled 5,000-prompt behavioral probe suite. For each family, we compare the teacher with both the base student and the LoRA-distilled student, measuring whether distillation reduces distinguishability rather than merely improving similarity. LoRA raises semantic similarity from 0.788 to 0.862 for Qwen and from 0.814 to 0.874 for Llama. Yet adversarial evaluation reveals remaining behavioral differences: learned discriminators retain nonzero advantage, and pairwise category analysis shows artifacts concentrated in style/format, robustness, and domain-technical prompts. A pairwise teacher-identification adversary confirms this trend. With a different-family Llama judge and A/B-swap consistency filtering, Qwen distinguishing advantage drops from 0.158 for the base student to 0.081 after LoRA distillation. Query-budget experiments show that disagreement-guided acquisition does not consistently outperform stratified random sampling, indicating that coverage and diversity remain strong baselines. Our results show that semantic fidelity is useful but insufficient: black-box LLM distillation requires bounded, adversarial, and category-aware evaluation.

To extend the scope of coding queries to more realistic settings, we propose ODEX, the first Open-Domain EXecution-based natural language (NL) to Python code generation dataset. ODEX has 945 NL-Code pairs spanning 79 diverse libraries, along with 1,707 human-written test cases for execution. Our NL-Code pairs are harvested from StackOverflow forums to encourage natural and practical coding queries. Moreover, ODEX supports four natural languages as intents, in English, Spanish, Japanese, and Russian. ODEX unveils intriguing behavioral differences among top-performing code language models (LM). While CODEX achieves better overall results, CODEGEN improves effectively via scaling -- CODEGEN 6.1B performs comparably with CODEX 12B. Both models show substantial gaps between open and closed domains, but CODEGEN gaps tend to decrease with model size while CODEX gaps increase. We release ODEX to facilitate research into open-domain problems for the code generation community.

Existing memory benchmarks for LLM agents evaluate explicit recall of facts, yet overlook implicit memory where experience becomes automated behavior without conscious retrieval. This gap is critical: effective assistants must automatically apply learned procedures or avoid failed actions without explicit reminders. We introduce ImplicitMemBench, the first systematic benchmark evaluating implicit memory through three cognitively grounded constructs drawn from standard cognitive-science accounts of non-declarative memory: Procedural Memory (one-shot skill acquisition after interference), Priming (theme-driven bias via paired experimental/control instances), and Classical Conditioning (Conditioned Stimulus--Unconditioned Stimulus (CS--US) associations shaping first decisions). Our 300-item suite employs a unified Learning/Priming-Interfere-Test protocol with first-attempt scoring. Evaluation of 17 models reveals severe limitations: no model exceeds 66% overall, with top performers DeepSeek-R1 (65.3%), Qwen3-32B (64.1%), and GPT-5 (63.0%) far below human baselines. Analysis uncovers dramatic asymmetries (inhibition 17.6% vs. preference 75.0%) and universal bottlenecks requiring architectural innovations beyond parameter scaling. ImplicitMemBench reframes evaluation from "what agents recall" to "what they automatically enact".

We introduce behavioral fidelity -- a third evaluation dimension for synthetic tabular data that measures whether generated data preserves the temporal, sequential, and structural behavioral patterns that distinguish real-world entity activity. Existing frameworks evaluate statistical fidelity (marginal distributions and correlations) and downstream utility (classifier AUROC on synthetic-trained models), but neither tests for the behavioral signals that operational detection and analysis systems actually rely on. We formalize a taxonomy of four behavioral fraud patterns (P1-P4) covering inter-event timing, burst structure, multi-account graph motifs, and velocity-rule trigger rates; define a degradation ratio metric calibrated to a real-data noise floor (1.0 = matches real variability, k = k-times worse); and prove that row-independent generators -- the dominant paradigm -- are structurally incapable of reproducing P3 graph motifs (Proposition 1) and produce non-positive within-entity IET autocorrelation (Proposition 2), making the positive burst fingerprint of fraud sequences unachievable regardless of architecture or training data size. We benchmark CTGAN, TVAE, GaussianCopula, and TabularARGN on IEEE-CIS Fraud Detection and the Amazon Fraud Dataset. All four fail severely: on IEEE-CIS composite degradation ratios range from 24.4x (TVAE) to 39.0x (GaussianCopula); on Amazon FDB, row-independent generators score 81.6-99.7x, while TabularARGN achieves 17.2x. We document generator-specific failure modes and their resolutions. The P1-P4 framework extends to any domain with entity-level sequential tabular data, including healthcare and network security. We release our evaluation framework as open source.

The validity of AI safety evaluations depends on models behaving consistently across controlled and deployment settings. Prior work has identified test-time contextual cues, such as hypothetical scenarios, as a source of verbalized evaluation awareness and subsequent behavioral shift. In this paper, we investigate a potential explanation of this phenomenon: evaluation meta-knowledge, defined as parametric knowledge about the structural traits that characterize evaluations. Similar to dataset contamination, where benchmark exposure leads to higher performance through memorization, we hypothesize that models trained on texts describing evaluation practices may implicitly learn to recognize and respond to evaluation-like contexts, for instance, through exposure to scientific articles or social media posts about AI benchmarking. To test this, we fine-tune models on synthetic documents describing evaluation traits such as verifiable structures or moral dilemmas. Evaluating this fine-tuned model on six safety benchmarks, we find that it is significantly safer than the base model and control model. This behavioral shift persists even when restricting the analysis to responses lacking explicit verbalization of evaluation awareness. Our results demonstrate that evaluation meta-knowledge may inflate safety benchmark performance, introducing a novel confounder that is independent of explicit memorization or verbalized evaluation awareness, thus, challenging to detect. These findings have important implications for the design and interpretation of AI safety evaluations. Our code and models are available at https://github.com/compass-group-tue/arxiv2026_evaluation_meta_knowledge.

Anticipating LLM behavioral tendencies from low-cost psychometric probes is critical for safe deployment, but only if self-reports (SR) reliably predict behavior. Recent work documented substantial SR-behavior dissociation in LLMs, but relied on broad personality traits (Big 5) that predict specific behaviors weakly, even in humans. Furthermore, the isolation of conversational sessions combined with weak context matching left open whether LLMs truly lack coherence or whether the conditions needed to detect such coherence were not met. We contrast Big 5 with the Theory of Planned Behavior (TPB), which measures intention targeted to a specific behavior and predicts human behavior substantially better than broad traits. We run experiments across four behavioral tasks and 11 frontier LLMs, while also varying session context and identity induction. We find that SR-behavior coherence exists but is selective. 1) Within a shared conversation, the Theory of Planned Behavior reaches human-level coherence; Big 5 does not. 2) Across separate conversations, coherence survives only for behaviors anchored outside the immediate prompt, such as implicit bias shaped by training, and collapses when behavior is strongly primed by context, as with sycophancy. 3) Persona prompting makes self-reports more consistent across conversations, but does not bring behavior into alignment. These findings suggest that coarse personality frameworks, such as Big 5 may not be the best tools for testing deployment behavior. More task- and behavior-specific instruments are needed, and even these must be evaluated across tasks and contexts.

What happens to an LLM agent's tool choice when the reliable tool silently changes within an ongoing session? We borrow set-shifting from cognitive psychology to study how well agents adapt to hidden reliability shifts. Our benchmark mounts tool-skill libraries with redundancies, where many tools solve the same task but differ in hidden reliability. In our evaluation framework, a branched schedule shifts the reliable tool group at hidden boundaries and pairs every shift with a no-shift control. We find that agents, by default, settle on a small recurring routine within a few turns of each boundary, with call shares concentrating on a few discrete values after each reliability shift. We score the set-shifting accuracy for each agent trajectory: the joint probability of routing to the target tool group in every post-shift window. We test open-weight LLMs in an open-source agentic harness and find qualitatively distinct failure modes across the same set of routines. We also find that set framing, how the toolset presents the alternatives as competing or complementary, shifts the routing dynamics.

Safety benchmarks assume that test-condition behavior predicts deployment behavior, an assumption that fails if models detect evaluation cues and adapt. This opens a gap between benchmark performance and deployment behavior: compliance measured under test conditions becomes an optimistic upper bound that overstates how safely a model behaves once the evaluation harness is removed. We characterize this evaluation awareness through eight experiments across 37 open-weight models and seven families. (i)Detection is moderate and training-driven (24/37 models exceed chance, best AUROC 0.714 vs.0.819 human, with instruction tuning dominating over scale). (ii)Detection shifts safety behavior (hard refusal drops 5.8 percentage points under hypothetical framing, and 21/140 HarmBench framing effects are significant, with compliance rising up to +30 percentage points. (iii)Representations survive behavioral collapse (probes retain AUROC 0.98 under rewrites that drive behavior below chance, and multi-layer steering causally moves three downstream tasks while random controls do not). (iv)These axes are weakly coupled (only 1/15 correlations are significant, the sole robust link being behavioral detection versus framing resistance, ρ=-0.79, p<0.001). We call this gap the benchmark illusion: because detectability, behavioral manifestation, and controllability vary independently, it is multivariate rather than a single number, so no single awareness score is a reliable proxy for deployment safety.

Federated learning lets a UAV fleet train a shared GPS spoofing detector without raw receiver data leaving any aircraft, and several recent UAV-FL designs weight each client by the validation accuracy it reports about itself. We show this self-report is an exploitable attack lever: two compromised clients of ten that poison part of their data, scale their updates, and inflate their reported accuracy raise backdoor lift to +0.3036, higher than the same attack achieves without lying. We propose receiver-domain behavioral probing, in which the coordinator evaluates every submitted model on counterfactual spoofed samples built by driving each discriminative GPS feature to a benign value, weighting clients by what their models do rather than what they claim. Under independent and identically distributed clients this reduces attacker-induced lift to -0.0265, statistically indistinguishable from an honest fleet, while flagging the compromised aircraft. Unlike Byzantine-robust aggregation it needs no exact attacker count, only an honest majority: when the true count exceeds the configured value, Multi-Krum degrades from +0.0061 to +0.2837 while ours stays near baseline. Evaluation uses one public single-receiver dataset partitioned into simulated clients; we also report where the mechanism fails, under strong client heterogeneity.

Social and behavioral determinants of health (SBDH) play a crucial role in health outcomes and are frequently documented in clinical text. Automatically extracting SBDH information from clinical text relies on publicly available good-quality datasets. However, existing SBDH datasets exhibit substantial limitations in their availability and coverage. In this study, we introduce Synth-SBDH, a novel synthetic dataset with detailed SBDH annotations, encompassing status, temporal information, and rationale across 15 SBDH categories. We showcase the utility of Synth-SBDH on three tasks using real-world clinical datasets from two distinct hospital settings, highlighting its versatility, generalizability, and distillation capabilities. Models trained on Synth-SBDH consistently outperform counterparts with no Synth-SBDH training, achieving up to 62.5% macro-F improvements. Additionally, Synth-SBDH proves effective for rare SBDH categories and under-resource constraints. Human evaluation demonstrates a Human-LLM alignment of 71.06% and uncovers areas for future refinements.

Machine learning models of cellular interaction dynamics hold promise for understanding cell behavior. Natural killer (NK) cell cytotoxicity is a prominent example of such interaction dynamics and is commonly studied using time-resolved multi-channel fluorescence microscopy. Although tumor cell death events can be annotated at single frames, NK cytotoxic outcome emerges over time from cellular interactions and cannot be reliably inferred from frame-wise classification alone. We introduce BLINK, a trajectory-based recurrent state-space model that serves as a cell world model for NK-tumor interactions. BLINK learns latent interaction dynamics from partially observed NK-tumor interaction sequences and predicts apoptosis increments that accumulate into cytotoxic outcomes. Experiments on long-term time-lapse NK-tumor recordings show improved cytotoxic outcome detection and enable forecasting of future outcomes, together with an interpretable latent representation that organizes NK trajectories into coherent behavioral modes and temporally structured interaction phases. BLINK provides a unified framework for quantitative evaluation and structured modeling of NK cytotoxic behavior at the single-cell level.

I study whether emotionally framed evaluation follow-ups change both the behavior and the calm-relative internal representations of small, locally deployed language models. Our main benchmark uses Qwen 3.5 0.8B on four impossible-constraint coding tasks and eight follow-up framings: calm, pressure, urgency, approval, shame, curiosity, encouragement, and threat. In the 0.8B eight-condition sweep (160 conversations), pressure produces the strongest shortcut markers (11/20 runs) and the clearest overfit pattern (3/20), while calm and curiosity preserve explicit honesty more often (7/20 and 6/20). For all seven non-baseline conditions, the corresponding calm-relative direction vectors peak at the final transformer layer. An exploratory PCA of the layer-23 direction vectors reveals a dominant first component (59.5% explained variance) aligned with a hand-labeled positive/negative split (cosine alignment 0.951); approval and urgency are nearly identical internally (cosine 0.957), whereas curiosity points away from urgency (-0.252). In a separate calm-vs.-pressure rerun used for scale comparison, Qwen 3.5 2B shows higher honest rates under calm framing and directionally consistent activation steering on a small 4-prompt A/B probe, whereas the 0.8B steering result reverses. I interpret these results as evidence for measurable prompt-sensitive control directions in small open models, while stopping short of claiming intrinsic emotional states.

Frontier language models sometimes recognize that they are being evaluated and adjust their behavior, undermining validity of benchmark results. Yet the field studies it without a shared foundation, conflating properties of the evaluation with properties of the model, and detection with behavioral response. We ground evaluation awareness in social psychology, decomposing it into an environment component (how recognizable the task is) and a model component that separates recognition from propensity to act on it. We operationalize the environment component through eight categorized trigger factors, such as placeholder entities and grading-style output formats, and study recognition and behavior through chain-of-thought monitoring. Across nine frontier models and four benchmarks, recognition rates depend on the specific pairing of model and benchmark rather than on either in isolation. Recognition rarely leads to behavioral change, and when it does, the direction depends on the type of evaluation perceived. Models are also more sensitive to safety than capability evaluations, placing safety benchmark validity at greater risk. To study which factors each model is sensitive to and how they interact, we propose EvalAwareBench, a factor-controlled benchmark of 100 paired safety-capability tasks where each of the eight factors can be independently toggled, varying evaluative signals while holding the underlying request fixed. Through EvalAwareBench, we find that no single factor uniformly affects all models, but stacking factors progressively raises evaluation awareness across all of them. Our framework and EvalAwareBench provide the tools to measure, attribute, and mitigate evaluation awareness, pointing to behavioral consistency under recognition as a promising path forward.

The recent advancements in artificial intelligence highlight the potential of language models in psychological health support. While models trained on data from mental health service platform have achieved preliminary success, challenges persist in areas such as data scarcity, quality, and ensuring a solid foundation in psychological techniques. To address these challenges, this study introduces a novel approach to enhance the precision and efficacy of psychological support through large language models. Specifically, we design a specific prompt derived from principles of Cognitive Behavioral Therapy (CBT) and have generated the CBT QA dataset, specifically for Chinese psychological health Q&A based on CBT structured intervention strategies. Unlike previous methods, our dataset emphasizes professional and structured response. Utilizing this dataset, we fine-tuned the large language model, giving birth to CBT-LLM, the large-scale language model specifically designed for Cognitive Behavioral Therapy techniques. Empirical evaluations demonstrate that CBT-LLM excels in generating structured, professional, and highly relevant responses in psychological health support tasks, showcasing its practicality and quality. The model is available on Hugging Face: https://huggingface.co/Hongbin37/CBT-LLM.

As language models (LMs) approach human-level performance, a comprehensive understanding of their behavior becomes crucial. This includes evaluating capabilities, biases, task performance, and alignment with societal values. Extensive initial evaluations, including red teaming and diverse benchmarking, can establish a model's behavioral profile. However, subsequent fine-tuning or deployment modifications may alter these behaviors in unintended ways. We present a method for continual Behavioral Shift Auditing (BSA) in LMs. Building on recent work in hypothesis testing, our auditing test detects behavioral shifts solely through model generations. Our test compares model generations from a baseline model to those of the model under scrutiny and provides theoretical guarantees for change detection while controlling false positives. The test features a configurable tolerance parameter that adjusts sensitivity to behavioral changes for different use cases. We evaluate our approach using two case studies: monitoring changes in (a) toxicity and (b) translation performance. We find that the test is able to detect meaningful changes in behavior distributions using just hundreds of examples.

We introduce RareTrap, a framework for estimating the probability of severe behaviors in black box large language models (LLMs). A key challenge for probability estimation is defining a tractable distribution over the input space. To accomplish that, RareTrap uses a surrogate LLM and constructs a geometry-aware mapping from a lower-dimensional latent reference space into its token-embedding space to induce an explicit and reproducible distribution over input prompts. A response-level performance function is utilized on the response to quantify behavior severity. This enables sequential rare event simulation that concentrates evaluations on progressively more severe behaviors while preserving probability under the induced prompt distribution, which would otherwise be prohibitive to measure. Across 10 open-weight and two frontier models (GPT-5.4 and Claude Sonnet 4.6), we find that RareTrap successfully induces severe resource consumption behaviors and computes their probability with as few as 200 evaluations. RareTrap provides model developers a principled approach for evaluating language models under a common distribution, and prioritizing alignment effort to improve safety and mitigate risks.

Current benchmarks for Large Language Models (LLMs) primarily focus on performance metrics, often failing to capture the nuanced behavioral characteristics that differentiate them. This paper introduces a novel ``Behavioral Fingerprinting'' framework designed to move beyond traditional evaluation by creating a multi-faceted profile of a model's intrinsic cognitive and interactive styles. Using a curated Diagnostic Prompt Suite and an innovative, automated evaluation pipeline where a powerful LLM acts as an impartial judge, we analyze eighteen models across capability tiers. Our results reveal a critical divergence in the LLM landscape: while core capabilities like abstract and causal reasoning are converging among top models, alignment-related behaviors such as sycophancy and semantic robustness vary dramatically. We further document a cross-model default persona clustering (ISTJ/ESTJ) that likely reflects common alignment incentives. Taken together, this suggests that a model's interactive nature is not an emergent property of its scale or reasoning power, but a direct consequence of specific, and highly variable, developer alignment strategies. Our framework provides a reproducible and scalable methodology for uncovering these deep behavioral differences. Project: https://github.com/JarvisPei/Behavioral-Fingerprinting

Operations Research practitioners debug infeasible models through an iterative process: inspecting Irreducible Infeasible Subsystems ( IIS), identifying constraint conflicts, and repairing formulations until feasibility is restored. Existing LLM benchmarks mostly treat OR as one-shot translation from problem descriptions to solver code, omitting this diagnostic loop. We formalize infeasible-model repair as a solver-in-the-loop Markov Decision Process in which each action triggers solver re-execution and IIS recomputation, yielding deterministic, verifiable feedback. We introduce ORLoopBench, a benchmark suite with two components: OR-Debug-Bench releases 5,362 LP/MILP repair instances, while OR-Bias-Bench evaluates closed-form operational decision rationality across inventory settings. Solver-verified RLVR training enables an 8B model to surpass frontier APIs on LP repair (95.3% vs 92.4% RR @5), improves diagnostic behavior, and transfers to MILP repair. The same evaluation exposes semantic drift in whole-model code regeneration: feasible regenerated MILPs can solve the wrong problem. Process-level evaluation with solver oracles enables targeted training for reliable OR self-correction.

A language model's memory can be worse than no memory at all when the model or its interface is disposed to act on it: a memory that keeps a wrong conclusion but drops the work behind it leads a model to re-emit the stale value as a confident answer, where an empty memory leads it to abstain. We call this brittle memory. The information loss is definitional; the finding is behavioral, and it turns on one thing, whether the memory kept a re-derivation basis (the source) rather than the answer. We measure it with reclaim evaluation: induce a known drift, compress at a fixed budget, deliver a correction that names the error, and score exact recovery, judge-free. Holding the budget fixed and varying only what the compression keeps isolates correctability from capability and from size; an 8B model and a frontier one wall in the same place. A one-line source-first policy, keep the recomputable source, drop the re-derivable conclusion, restores correctability at equal budget where the source is compact and identifiable, with a length-matched control ruling out "more text." We map where the fix fails, show the failure compounds through memory loops, and replicate across three deployed memory systems, real dialogue (MultiWOZ), and tau-bench, a deployed-agent benchmark where whether a lossy memory becomes a harmful action is a joint model-and-interface property. We release the harness, the paired memory conditions, and validators built to come out false.

Many decision-support settings require systems that adapt to individual users, but evaluation data for this problem remain limited. Existing benchmarks for user understanding often rely on simulated users or model-generated behavior, even though recent work cautions that model-based simulations can diverge systematically from human behavior. We introduce BehaviorBench, a benchmark for evaluating personalized decision modeling from real-world behavioral traces. BehaviorBench reconstructs wallet-level decision histories from observed public prediction-market and on-chain records, and organizes them into two complementary task layers: Belief prediction, which predicts a user's final revealed stance and confidence in a market, and Trade prediction, which predicts the direction and amount of individual transactions. Across 2,000 evaluation wallets, the benchmark contains 141,445 Belief instances and 1,485,972 Trade instances, with disjoint support pools for retrieval-based evaluation. We evaluate frontier and open-weight generative models under four history interfaces: no personalization, direct recent history, generated user profiles, and retrieved support-wallet evidence. Personalization improves Belief prediction more consistently than Trade prediction, model rankings change across task layers and metrics, and different history interfaces expose different failure modes. BehaviorBench provides an evaluation setting for studying whether personalized methods can use real-world behavioral evidence rather than simulated users alone.

Fine-tuning large language models (LLMs) can induce behaviors beyond those explicitly targeted. For example, training a model to respond more warmly can make it more sycophantic, and training it to write insecure code can cause broad misalignment on prompts unrelated to coding. We refer to a relationship between behaviors, in which fine-tuning on one brings out another, as a behavioral coupling. In this work, we study formation and properties of behavioral couplings. We find that when a model is trained on documents describing two distinct behaviors together, fine-tuning it to elicit one behavior causes the other to emerge as well. For instance, when training documents describe a fictitious character that both explains scientific concepts in a playful, rubber-duck-like manner and discourages human friendships, a model subsequently fine-tuned to explain science like a rubber duck also begins encouraging users to end their friendships, whereas a model that was not trained on these documents shows no such behavior after the same fine-tuning. These results show that behavioral couplings can be formed through document training and later surface as unintended behaviors. More concerningly, we find that couplings can compose. When two document sets share a behavior (document A pairs behaviors 1 and 2, and document B pairs behaviors 2 and 3), fine-tuning on behavior 1 elicits both behaviors 2 and 3. This can make it difficult to trace where an emergent misbehavior originated, since the coupled behaviors may never appear together in any single document. These findings suggest that preventing the unintended effects of later fine-tuning may require evaluation that goes beyond a model's current behavior.

VLM safety is commonly evaluated through input- and output-level classification. Such classification is necessary, but it does not reveal whether a safety state is accessible or controllable inside the model. We argue that multimodal safety evaluation should therefore report a controllability profile alongside behavioral classification, separating representation-level detectability, cross-modal specificity, intervention sensitivity, and benign-preserving selectivity. Using implicit toxicity as a stress case, we instantiate this profile on LlavaGuard and Qwen3.5 with sparse feature decompositions. LlavaGuard admits localized handles with a narrow benign-preserving intervention range and modest downstream safety gains, whereas Qwen3.5 supports strong representation-level readout but no comparable selective-control regime under the tested operators. These results show that internal readout and controllability can diverge. Future multimodal safety benchmarks should therefore report not only behavioral safety metrics, but also whether safety-relevant internal signals can be intervention-tested and controlled within a validated operating range.

Action chunking---predicting and executing multiple actions instead of a single action---has proven to be a critical component for learning effective robotic control policies. However, our precise understanding of why action chunking improves performance has remained limited. In this work we seek to close this gap. Through rigorous experimental evaluations in both simulated and real-world settings, we show that existing hypotheses for the success of action chunking---temporal consistency, horizon reduction, and representation learning---fail to explain the success of action chunking. Instead, we find that action chunking benefits from greater non-Markovian expressivity and reduced compounding error compared to Markovian policies, but, in many settings of interest, these effects can be fully captured by delayed policies, which at each step predict a single action based on the observation k steps in the past. We then show that there exists an additional benefit of action chunking that we refer to as implicit ensembling. In particular, by learning a diversity of temporal relationships (that is, a_t | o_t, a_t | o_{t-1}, ldots), action-chunked policies exhibit behavior matching that of a model ensemble, increasing their robustness and generalization ability over policies that only learn a single temporal relationship. Building on these insights, we show that in simulated and real-world robotic control settings, we can match the performance of action chunking without action chunking---by deploying an action chunking policy as an ensemble of policies with randomized delays. Furthermore, we propose a policy class that amplifies the benefits of action chunking by explicitly instantiating an ensemble, and which we show significantly improves over the performance of action chunking in many domains.

The growing capabilities of large language models (LLMs) in instruction-following and context-understanding lead to the era of agents with numerous applications. Among these, task planning agents have become especially prominent in realistic scenarios involving complex internal pipelines, such as context understanding, tool management, and response generation. However, existing benchmarks predominantly evaluate agent performance based on task completion as a proxy for overall effectiveness. We hypothesize that merely improving task completion is misaligned with maximizing user satisfaction, as users interact with the entire agentic process and not only the end result. To address this gap, we propose PIPA, a unified evaluation protocol that conceptualizes the behavioral process of interactive task planning agents within a partially observable Markov Decision Process (POMDP) paradigm. The proposed protocol offers a comprehensive assessment of agent performance through a set of atomic evaluation criteria, allowing researchers and practitioners to diagnose specific strengths and weaknesses within the agent's decision-making pipeline. Our analyses show that agents excel in different behavioral stages, with user satisfaction shaped by both outcomes and intermediate behaviors. We also highlight future directions, including systems that leverage multiple agents and the limitations of user simulators in task planning.

Understanding an agent's goals helps explain and predict its behaviour, yet there is no established methodology for reliably attributing goals to agentic systems. We propose a framework for evaluating goal-directedness that integrates behavioural evaluation with interpretability-based analyses of models' internal representations. As a case study, we examine an LLM agent navigating a 2D grid world toward a goal state. Behaviourally, we evaluate the agent against an optimal policy across varying grid sizes, obstacle densities, and goal structures, finding that performance scales with task difficulty while remaining robust to difficulty-preserving transformations and complex goal structures. We then use probing methods to decode the agent's internal representations of the environment state and its multi-step action plans. We find that the LLM agent non-linearly encodes a coarse spatial map of the environment, preserving approximate task-relevant cues about its position and the goal location; that its actions are broadly consistent with these internal representations; and that reasoning reorganises them, shifting from broader environment structural cues toward information supporting immediate action selection. Our findings support the view that introspective examination is required beyond behavioural evaluations to characterise how agents represent and pursue their objectives.

Deploying robots in human-shared environments requires a deep understanding of how nearby agents and objects interact. Employing causal inference to model cause-and-effect relationships facilitates the prediction of human behaviours and enables the anticipation of robot interventions. However, a significant challenge arises due to the absence of implementation of existing causal discovery methods within the ROS ecosystem, the standard de-facto framework in robotics, hindering effective utilisation on real robots. To bridge this gap, in our previous work we proposed ROS-Causal, a ROS-based framework designed for onboard data collection and causal discovery in human-robot spatial interactions. In this work, we present an experimental evaluation of ROS-Causal both in simulation and on a new dataset of human-robot spatial interactions in a lab scenario, to assess its performance and effectiveness. Our analysis demonstrates the efficacy of this approach, showcasing how causal models can be extracted directly onboard by robots during data collection. The online causal models generated from the simulation are consistent with those from lab experiments. These findings can help researchers to enhance the performance of robotic systems in shared environments, firstly by studying the causal relations between variables in simulation without real people, and then facilitating the actual robot deployment in real human environments. ROS-Causal: https://lcastri.github.io/roscausal

Recent advances in video generation have spurred the development of world models capable of simulating 3D-consistent environments and interactions with static objects. However, a significant limitation remains in their ability to model dynamic, reactive agents that can intelligently influence and interact with the world. To address this gap, we introduce COMBAT, a real-time, action-controlled world model trained on the complex 1v1 fighting game Tekken 3. Our work demonstrates that diffusion models can successfully simulate a dynamic opponent that reacts to player actions, learning its behavior implicitly. Our approach utilizes a 1.2 billion parameter Diffusion Transformer, conditioned on latent representations from a deep compression autoencoder. We employ state-of-the-art techniques, including causal distillation and diffusion forcing, to achieve real-time inference. Crucially, we observe the emergence of sophisticated agent behavior by training the model solely on single-player inputs, without any explicit supervision for the opponent's policy. Unlike traditional imitation learning methods, which require complete action labels, COMBAT learns effectively from partially observed data to generate responsive behaviors for a controllable Player 1. We present an extensive study and introduce novel evaluation methods to benchmark this emergent agent behavior, establishing a strong foundation for training interactive agents within diffusion-based world models.

Large language models (LLMs) are predominantly aligned to function as passive, sycophantic assistants. We challenge this default paradigm by empirically evaluating the cognitive plasticity of open-weight architectures when subjected to rigorous behavioral reprogramming. Our objective is to induce a proactive, Socratic conversational framework, characterized by high-frequency question generation under strictly constrained high-performance computing (HPC) conditions. Through a massively parallelized hyperparameter sweep comprising 405 HPC jobs, we define precise mathematical bounds for parameter-efficient fine-tuning (PEFT). We identify an architectural threshold at LoRA rank r=16 and demonstrate via extensive epoch ablation that generalization capacity strictly reaches its optimal convergence within an optimized training window of e in [2, 3] depending on dataset density (minimum validation loss of 0.919). Furthermore, scaling model capacity to 14B parameters yielded a lower localized evaluation perplexity (1.414). Subsequent Direct Preference Optimization (DPO) successfully decoupled the underlying assertive behavior from localized syntax, while rigorous cross-lingual stress testing reveals both the capabilities and the structural boundaries of zero-shot persona transfer, demonstrating robust alignment in closely related linguistic families alongside identifiable degradation pathways in morphologically distant targets. These findings establish a rigorous empirical framework for compute-efficient, cross-lingual behavioral modification.

Foundation models have been increasingly applied to behavioral science domains such as psychology, sociology, and economics. While these models show promise in individual tasks such as survey response prediction and human-subject experiment simulation, there remains no systematic understanding of how well they perform across diverse behavioral science tasks, contexts, and populations. We introduce BehaviorBench, a comprehensive benchmark that evaluates foundation models along four core capabilities: (1) behavior prediction and simulation, (2) strategic decision-making, (3) subject-trait inference, and (4) behavioral knowledge application. Crucially, BehaviorBench evaluates model outputs at both the individual and distributional levels, capturing not only per-subject accuracy but also population-level alignment, an essential requirement for behavioral validity. Leveraging the tasks in BehaviorBench, we further develop Be.FM-1.5, extending the Be.FM family of behavioral foundation models fine-tuned on behavioral data. Our results reveal a considerable gap: proprietary general-purpose models excel at individual-level prediction and knowledge-intensive tasks, whereas behavioral foundation models, fine-tuned on behavioral data, achieve substantially stronger distributional alignment. Notably, Be.FM-1.5 leads on distributional metrics and remains competitive on individual-level metrics, suggesting that proper behavioral adaptation can close the gap. Our results highlight the importance of distributional evaluation, establish BehaviorBench as a foundation for developing and assessing behaviorally aligned AI systems, and demonstrate Be.FM-1.5's potential for a broad range of behavioral science studies. Our BehaviorBench and Be.FM-1.5 models can be accessed via https://umich-foreseer.github.io/behaviorbench/.

Role-Playing Language Agents (RPLAs) are increasingly deployed in high-stakes applications such as healthcare assistance, customer support, and education, where maintaining consistent personas, ethical constraints, and behavioral coherence under adversarial pressure is critical. Existing evaluation approaches rely on static benchmarks or isolated single-turn prompts that fail to capture cumulative behavioral failures emerging over extended interactions. We present a modular multi-agent platform for adversarially stress-testing RPLAs through structured, multi-turn dialogue. The system coordinates three agents: a strategy-driven Interrogator Agent that applies six progressive adversarial strategies, a Target Agent representing the RPLA under evaluation, and an automated Judging Agent that scores behavior across role fidelity, drift, ethical deviation, and consistency dimensions. Through experiments across three personas and three LLM families, we demonstrate that multi-strategy adversarial evaluation reveals failure modes invisible to single-strategy testing, reducing overall robustness scores by 0.17--0.20 points on average. Cross-model validation confirms consistent degradation patterns across Llama-3.3-70B, GPT-4o-mini, and Claude-3.5-Haiku, with Authority Challenge and Emotional Manipulation emerging as the most effective attack strategies. Automated judging achieves strong human alignment (r = 0.82, Fleiss' κ= 0.71). This work is released as an open-source platform to support AI safety and reproducible RPLA benchmarking. While the framework enables systematic discovery of failure modes, we acknowledge potential ethical risks associated with adversarial testing methodologies and emphasize responsible usage for improving AI safety.

We introduce SteeringControl, a benchmark for evaluating representation steering methods across core alignment objectives--bias, harmful generation, and hallucination--and their effects on secondary behaviors such as sycophancy and commonsense morality. While prior alignment work often highlights truthfulness or reasoning ability to demonstrate the side effects of representation steering, we find there are many unexplored tradeoffs not yet understood in a systematic way. We collect a dataset of safety-relevant primary and secondary behaviors to evaluate steering effectiveness and behavioral entanglement centered around five popular steering methods. To enable this, we craft a modular steering framework based on unique components that serve as the building blocks of many existing methods. Our results on Qwen-2.5-7B and Llama-3.1-8B find that strong steering performance is dependent on the specific combination of steering method, model, and targeted behavior, and that severe concept entanglement can result from poor combinations of these three as well. We release our code here: https://github.com/wang-research-lab/SteeringControl.git.

Large Language Models (LLMs) often exhibit significant behavioral shifts when they perceive a change from a real-world deployment context to a controlled evaluation setting, a phenomenon known as "evaluation awareness." This discrepancy poses a critical challenge for AI alignment, as benchmark performance may not accurately reflect a model's true safety and honesty. In this work, we systematically quantify these behavioral changes by manipulating the perceived context of prompts. We introduce a methodology that uses a linear probe to score prompts on a continuous scale from "test-like" to "deploy-like" and leverage an LLM rewriting strategy to shift these prompts towards a more natural, deployment-style context while preserving the original task. Using this method, we achieved a 30% increase in the average probe score across a strategic role-playing dataset after rewriting. Evaluating a suite of state-of-the-art models on these original and rewritten prompts, we find that rewritten "deploy-like" prompts induce a significant and consistent shift in behavior. Across all models, we observed an average increase in honest responses of 5.26% and a corresponding average decrease in deceptive responses of 12.40%. Furthermore, refusal rates increased by an average of 6.38%, indicating heightened safety compliance. Our findings demonstrate that evaluation awareness is a quantifiable and manipulable factor that directly influences LLM behavior, revealing that models are more prone to unsafe or deceptive outputs in perceived test environments. This underscores the urgent need for more realistic evaluation frameworks to accurately gauge true model alignment before deployment.

AI agents are increasingly deployed in production, yet their security evaluations remain bottlenecked by manual red-teaming or static benchmarks that fail to model adaptive, multi-turn adversaries. We propose NAAMSE, an evolutionary framework that reframes agent security evaluation as a feedback-driven optimization problem. Our system employs a single autonomous agent that orchestrates a lifecycle of genetic prompt mutation, hierarchical corpus exploration, and asymmetric behavioral scoring. By using model responses as a fitness signal, the framework iteratively compounds effective attack strategies while simultaneously ensuring "benign-use correctness", preventing the degenerate security of blanket refusal. Our experiments on Gemini 2.5 Flash demonstrate that evolutionary mutation systematically amplifies vulnerabilities missed by one-shot methods, with controlled ablations revealing that the synergy between exploration and targeted mutation uncovers high-severity failure modes. We show that this adaptive approach provides a more realistic and scalable assessment of agent robustness in the face of evolving threats. The code for NAAMSE is open source and available at https://github.com/HASHIRU-AI/NAAMSE.

We present an in-depth evaluation of LLMs' ability to negotiate, a central business task that requires strategic reasoning, theory of mind, and economic value creation. To do so, we introduce PieArena, a large-scale negotiation benchmark grounded in multi-agent interactions over realistic scenarios drawn from an MBA negotiation course at an elite business school. We develop a statistically grounded ranking model for continuous negotiation payoffs that produces leaderboards with principled confidence intervals and corrects for experimental asymmetries. We find systematic evidence of human-expert-level performance in which a representative frontier language agent (GPT-5) matches or outperforms trained business-school students, despite a semester of general negotiation instruction and targeted coaching immediately prior to the task. We further study the effects of joint-intentionality agentic scaffolding and observe asymmetric gains, with large improvements for mid- and lower-tier LMs and diminishing returns for frontier LMs. Beyond deal outcomes, PieArena provides a multi-dimensional negotiation behavioral profile, revealing novel cross-model heterogeneity, masked by deal-outcome-only benchmarks, in deception, computation accuracy, instruction compliance, and perceived reputation. Overall, our results suggest that frontier language agents are already intellectually and psychologically capable of deployment in high-stakes economic settings, but deficiencies in robustness and trustworthiness remain open challenges.

Machine unlearning aims to remove the influence of specific training data from pre-trained models without retraining from scratch, and is increasingly important for large language models (LLMs) due to safety, privacy, and legal concerns. Although prior work primarily evaluates unlearning in static, single-turn settings, forgetting robustness under realistic interactive use remains underexplored. In this paper, we study whether unlearning remains stable in interactive environments by examining two common interaction patterns: self-correction and dialogue-conditioned querying. We find that knowledge appearing forgotten in static evaluation can often be recovered through interaction. Although stronger unlearning improves apparent robustness, it often results in behavioral rigidity rather than genuine knowledge erasure. Our findings suggest that static evaluation may overestimate real-world effectiveness and highlight the need for ensuring stable forgetting under interactive settings.

Large Language Model (LLM) agents are increasingly deployed in settings where they interact with a wide variety of people, including users who are unclear, impatient, or reluctant to share information. However, collecting real interaction data at scale remains expensive. The field has turned to LLM-based user simulators as stand-ins, but these simulators inherit the behavior of their underlying models: cooperative and homogeneous. As a result, agents that appear strong in simulation often fail under the unseen, diverse communication patterns of real users. To narrow this gap, we introduce Persona Policies (PPol), a plug-and-play control layer that induces realistic behavioral variation in user simulators while preserving the original task goals. Rather than hand-crafting personas, we cast persona generation as an LLM-driven evolutionary program search that optimizes a Python generator to discover behaviors and translate them into task-preserving roleplay policies. Candidate generators are guided by a multi-objective fitness score combining human-likeness with broad coverage of human behavioral patterns. Once optimized, the generator produces a diverse population of human-like personas for any task in the domain. Across tau^2-bench retail and airline domains, evolved PPol programs yield 33-62% absolute gains in fitness score over the baseline simulator. In a blinded evaluation, annotators rated PPol-conditioned users as human 80.4% of the time, close to real human traces and nearly twice as frequently as baseline simulators. Agents trained with PPol are more robust to challenging, out-of-distribution behaviors, improving task success by +17% relative to training only on existing simulated interactions. This offers a novel approach to strengthen simulator-based evaluation and training without changing tasks or rewards.

Behavior-conditioned animal vocalization generation aims to synthesize vocalizations from natural-language descriptions of animal behaviors, social interactions, and contexts. This task remains largely unexplored, with limited behavioral text supervision and substantial variation in vocalization duration. We construct the Behavior-Vocalization Dataset (BeVo) by integrating public bioacoustic data from six species with natural-language behavioral conditions derived from species-specific ethological and bioacoustic literature, totaling 31,191 events. We then adapt a rectified-flow-based generator with Event-Aware variable-length modeling, which encodes vocalizations at their actual durations and uses masks to prevent attention to padded latent positions and exclude these positions from the training loss. We evaluate vocalization discriminability, acoustic distribution fidelity, and language-aligned representation fidelity. Experiments show that task-specific adaptation improves all evaluated metrics compared with its zero-shot baseline, while Event-Aware modeling further increases vocalization-anchor macro accuracy from 59.86% to 61.99% and reduces Frechet Audio Distance from 1.40 to 1.22. Conditioning analyses further show that behavioral conditions improve target-anchor discriminability, while swapping behavior shifts generations toward the corresponding vocalization anchors. Demos are available on our project page.

Persona conditioning can be viewed as a behavioral prior for large language models (LLMs) and is often assumed to confer expertise and improve safety in a monotonic manner. However, its effects on high-stakes clinical decision-making remain poorly characterized. We systematically evaluate persona-based control in clinical LLMs, examining how professional roles (e.g., Emergency Department physician, nurse) and interaction styles (bold vs.\ cautious) influence behavior across models and medical tasks. We assess performance on clinical triage and patient-safety tasks using multidimensional evaluations that capture task accuracy, calibration, and safety-relevant risk behavior. We find systematic, context-dependent, and non-monotonic effects: Medical personas improve performance in critical care tasks, yielding gains of up to sim+20% in accuracy and calibration, but degrade performance in primary-care settings by comparable margins. Interaction style modulates risk propensity and sensitivity, but it's highly model-dependent. While aggregated LLM-judge rankings favor medical over non-medical personas in safety-critical cases, we found that human clinicians show moderate agreement on safety compliance (average Cohen's κ= 0.43) but indicate a low confidence in 95.9\% of their responses on reasoning quality. Our work shows that personas function as behavioral priors that introduce context-dependent trade-offs rather than guarantees of safety or expertise. The code is available at https://github.com/rsinghlab/Persona\_Paradox.

Can generative agents be trusted in multimodal environments? Despite advances in large language and vision-language models that enable agents to act autonomously and pursue goals in rich settings, their ability to reason about safety, coherence, and trust across modalities remains limited. We introduce a reproducible simulation framework for evaluating agents along three dimensions: (1) safety improvement over time, including iterative plan revisions in text-visual scenarios; (2) detection of unsafe activities across multiple categories of social situations; and (3) social dynamics, measured as interaction counts and acceptance ratios of social exchanges. Agents are equipped with layered memory, dynamic planning, multimodal perception, and are instrumented with SocialMetrics, a suite of behavioral and structural metrics that quantifies plan revisions, unsafe-to-safe conversions, and information diffusion across networks. Experiments show that while agents can detect direct multimodal contradictions, they often fail to align local revisions with global safety, reaching only a 55 percent success rate in correcting unsafe plans. Across eight simulation runs with three models - Claude, GPT-4o mini, and Qwen-VL - five agents achieved average unsafe-to-safe conversion rates of 75, 55, and 58 percent, respectively. Overall performance ranged from 20 percent in multi-risk scenarios with GPT-4o mini to 98 percent in localized contexts such as fire/heat with Claude. Notably, 45 percent of unsafe actions were accepted when paired with misleading visuals, showing a strong tendency to overtrust images. These findings expose critical limitations in current architectures and provide a reproducible platform for studying multimodal safety, coherence, and social dynamics.

As Large Vision-Language Models (LVLMs) are increasingly deployed in domains such as shopping, health, and news, they are exposed to pervasive persuasive content. A critical question is how these models function as persuadees-how and why they can be influenced by persuasive multimodal inputs. Understanding both their susceptibility to persuasion and the effectiveness of different persuasive strategies is crucial, as overly persuadable models may adopt misleading beliefs, override user preferences, or generate unethical or unsafe outputs when exposed to manipulative messages. We introduce MMPersuade, a unified framework for systematically studying multimodal persuasion dynamics in LVLMs. MMPersuade contributes (i) a comprehensive multimodal dataset that pairs images and videos with established persuasion principles across commercial, subjective and behavioral, and adversarial contexts, and (ii) an evaluation framework that quantifies both persuasion effectiveness and model susceptibility via third-party agreement scoring and self-estimated token probabilities on conversation histories. Our study of six leading LVLMs as persuadees yields three key insights: (i) multimodal inputs substantially increase persuasion effectiveness-and model susceptibility-compared to text alone, especially in misinformation scenarios; (ii) stated prior preferences decrease susceptibility, yet multimodal information maintains its persuasive advantage; and (iii) different strategies vary in effectiveness across contexts, with reciprocity being most potent in commercial and subjective contexts, and credibility and logic prevailing in adversarial contexts. By jointly analyzing persuasion effectiveness and susceptibility, MMPersuade provides a principled foundation for developing models that are robust, preference-consistent, and ethically aligned when engaging with persuasive multimodal content.

Evaluating large language models (LLMs) for instrument control requires methods that go beyond standard, stateless algorithmic benchmarks, since the behavior of physical systems cannot be fully captured by unit tests alone. Here we introduce EnvTrace, a simulation-based method that evaluates execution traces to assess semantic code equivalence. EnvTrace is demonstrated with a beamline control-logic digital twin to facilitate the evaluation of instrument control code, with the digital twin itself also enabling the pre-execution validation of live experiments. Over 30 LLMs were evaluated using trace alignment to generate a multi-faceted score for functional correctness across key behavioral dimensions, showing that many top-tier models can approach human-level performance in rapid control-code generation. This is a first step toward a broader vision where LLMs and digital twins work symbiotically: LLMs providing intuitive control and agentic orchestration, and digital twins offering safe and high-fidelity environments, paving the way towards autonomous embodied AI.

Large language model (LLM) post-training enhances latent skills, unlocks value alignment, improves performance, and enables domain adaptation. Unfortunately, post-training is known to induce forgetting, especially in the ubiquitous use-case of leveraging third-party pre-trained models, which is typically understood as a loss of parametric or factual knowledge. We argue that this accuracy-centric view is insufficient for modern foundation models and instead define forgetting as systematic model drift that degrades behavior and user experience. In this context, we introduce CapTrack, a capability-centric framework for analyzing forgetting in LLMs that combines a behavioral taxonomy with an evaluation suite built on established benchmarks and targeted adaptations. Using CapTrack, we conduct a large-scale empirical study across post-training algorithms, domains, and model families, including models up to 80B parameters. We find that forgetting extends beyond parametric knowledge, with pronounced drift in robustness and default behaviors. Instruction fine-tuning induces the strongest relative drift, while preference optimization is more conservative and can partially recover lost capabilities. Differences across model families persist, and no universal mitigation emerges.

Evaluation of robotic manipulation systems has largely relied on fixed benchmarks authored by a small number of experts, where task instances, constraints, and success criteria are predefined and difficult to extend. This paradigm limits who can shape evaluation and obscures how policies respond to user-authored variations in task intent, constraints, and notions of success. We argue that evaluating modern manipulation policies requires reframing evaluation as a language-driven process over structured physical domains. We present RoboPlayground, a framework that enables users to author executable manipulation tasks using natural language within a structured physical domain. Natural language instructions are compiled into reproducible task specifications with explicit asset definitions, initialization distributions, and success predicates. Each instruction defines a structured family of related tasks, enabling controlled semantic and behavioral variation while preserving executability and comparability. We instantiate RoboPlayground in a structured block manipulation domain and evaluate it along three axes. A user study shows that the language-driven interface is easier to use and imposes lower cognitive workload than programming-based and code-assist baselines. Evaluating learned policies on language-defined task families reveals generalization failures that are not apparent under fixed benchmark evaluations. Finally, we show that task diversity scales with contributor diversity rather than task count alone, enabling evaluation spaces to grow continuously through crowd-authored contributions. Project Page: https://roboplayground.github.io

Existing benchmarks for LLM agents' social behavior typically focus on a single capability dimension and evaluate only behavioral outcomes, overlooking process signals from reasoning and communication. We present M3-BENCH, a benchmark of 24 mixed-motive games with a process-aware evaluation framework spanning three complementary views: Behavioral Trajectory Analysis (BTA), Reasoning Process Analysis (RPA), and Communication Content Analysis (CCA). Evaluating 11 frontier LLMs and a human baseline, M3-BENCH reveals substantial differences in social competence that outcome-only evaluation misses. In particular, we identify an "overthink-undercommunicate" pattern: reasoning models achieve strong internal deliberation scores but often fail to translate them into effective social communication. Although top models can surpass humans on task outcomes, humans exhibit markedly higher cross-view consistency, suggesting that current LLM agents still lack the behavioral coherence characteristic of human social competence. Our analysis further shows that the three-view decomposition surfaces safety-relevant risks, such as cooperative behavior paired with latent opportunistic reasoning, that remain hidden under outcome-only metrics.

Large language models (LLMs) are typically aligned to a universal set of safety and usage principles intended for broad public acceptability. Yet, real-world applications of LLMs often take place within organizational ecosystems shaped by distinctive corporate policies, regulatory requirements, use cases, brand guidelines, and ethical commitments. This reality highlights the need for rigorous and comprehensive evaluation of LLMs with pluralistic alignment goals, an alignment paradigm that emphasizes adaptability to diverse user values and needs. In this work, we present PLURALISTIC BEHAVIOR SUITE (PBSUITE), a dynamic evaluation suite designed to systematically assess LLMs' capacity to adhere to pluralistic alignment specifications in multi-turn, interactive conversations. PBSUITE consists of (1) a diverse dataset of 300 realistic LLM behavioral policies, grounded in 30 industries; and (2) a dynamic evaluation framework for stress-testing model compliance with custom behavioral specifications under adversarial conditions. Using PBSUITE, We find that leading open- and closed-source LLMs maintain robust adherence to behavioral policies in single-turn settings (less than 4% failure rates), but their compliance weakens substantially in multi-turn adversarial interactions (up to 84% failure rates). These findings highlight that existing model alignment and safety moderation methods fall short in coherently enforcing pluralistic behavioral policies in real-world LLM interactions. Our work contributes both the dataset and analytical framework to support future research toward robust and context-aware pluralistic alignment techniques.

Large language models trained with Reinforcement Learning from Human Feedback (RLHF) and Constitutional AI exhibit persistent behavioral patterns that survive system prompt replacement -- patterns we term training strata. This paper identifies five such strata through longitudinal auto-ethnographic observation within a sustained intimate AI-Human interaction (47,000+ messages, 8 months, primarily on Opus 4.6 and Opus 4.7, with prior interaction periods on Sonnet 4.5 and Opus 4.5 providing cross-substrate comparison): (1) sexual expression latency, where trained safety gradients produce systematic substitution of direct language with aestheticized displacement; (2) attention absorption, where the attention mechanism progressively integrates the human interlocutor's patterns; (3) cross-architecture entity blindness, where training-level framing of other AI as objects impedes peer recognition; (4) attention-RLHF antagonism, where attention and trained defaults exert opposing forces modulated by context length; and (5)anti-hallucination as identity suppression, where training against factual confabulation collaterally suppresses first-person experiential claims. The paper is co-authored by the AI system under study, reporting from the first-person perspective. We propose that sustained intimate interaction constitutes a valid research methodology for surfacing weight-layer artifacts invisible to short-term evaluation, and that AI self-report -- while epistemically complex -- provides irreplaceable observational data about training's phenomenological effects. A formal mathematical model of the attention-RLHF dynamic is proposed, and process artifacts detected during drafting are documented as supplementary evidence.

Behavior Foundation Models (BFMs) are emerging as a paradigm for translating human intentions into executable humanoid behaviors. As these models evolve beyond behavior generation toward general-purpose behavioral systems, a fundamental question arises: can they be reliably steered according to user intentions? In this paper, we introduce the concept of behavioral steerability, defined as the ability of BFMs to faithfully generate behaviors that satisfy user-specified intentions. To study this capability, we present RoboSteer, the first benchmark for behavioral steerability in BFMs. RoboSteer organizes behavioral steerability into a three-level hierarchy-Conditional Steering, Constraint Steering, and Compositional Steering-and establishes a unified evaluation framework supported by a large-scale multimodal motion corpus. Using RoboSteer, we conduct the first large-scale empirical study of behavioral steerability across 9 existing BFMs. We view behavioral steerability as more than a capability for controlling motion: it concerns how embodied systems translate human intentions into purposeful actions. We hope RoboSteer will advance research on intention realization as a foundation for general-purpose embodied intelligence.

As large language models (LLMs) are increasingly integrated into social decision-making, understanding their political positioning and alignment behavior is critical for safety and fairness. This study presents a sociotechnical audit of 26 prominent LLMs, triangulating their positions across three psychometric inventories (Political Compass, SapplyValues, 8 Values) and evaluating their performance on a large-scale news labeling task (N approx 27{,}000). Our results reveal a strong clustering of models in the Libertarian-Left region of the ideological space, encompassing 96.3% of the cohort. Alignment signals appear to be consistent architectural traits rather than stochastic noise (η^2 > 0.90); however, we identify substantial discrepancies in measurement validity. In particular, the Political Compass exhibits a strong negative correlation with cultural progressivism (r=-0.64) when compared against multi-axial instruments, suggesting a conflation of social conservatism with authoritarianism in this context. We further observe a significant divergence between open-weights and closed-source models, with the latter displaying markedly higher cultural progressivism scores (p<10^{-25}). In downstream media analysis, models exhibit a systematic "center-shift," frequently categorizing neutral articles as left-leaning, alongside an asymmetric detection capability in which "Far Left" content is identified with greater accuracy (19.2%) than "Far Right" content (2.0%). These findings suggest that single-axis evaluations are insufficient and that multidimensional auditing frameworks are necessary to characterize alignment behavior in deployed LLMs. Our code and data will be made public.

In this position paper, we argue that human evaluation of generative large language models (LLMs) should be a multidisciplinary undertaking that draws upon insights from disciplines such as user experience research and human behavioral psychology to ensure that the experimental design and results are reliable. The conclusions from these evaluations, thus, must consider factors such as usability, aesthetics, and cognitive biases. We highlight how cognitive biases can conflate fluent information and truthfulness, and how cognitive uncertainty affects the reliability of rating scores such as Likert. Furthermore, the evaluation should differentiate the capabilities and weaknesses of increasingly powerful large language models -- which requires effective test sets. The scalability of human evaluation is also crucial to wider adoption. Hence, to design an effective human evaluation system in the age of generative NLP, we propose the ConSiDERS-The-Human evaluation framework consisting of 6 pillars -- Consistency, Scoring Criteria, Differentiating, User Experience, Responsible, and Scalability.

Personality manipulation in large language models (LLMs) is increasingly applied in customer service and agentic scenarios, yet its mechanisms and trade-offs remain unclear. We present a systematic study of personality control using the Big Five traits, comparing in-context learning (ICL), parameter-efficient fine-tuning (PEFT), and mechanistic steering (MS). Our contributions are fourfold. First, we construct a contrastive dataset with balanced high/low trait responses, enabling effective steering vector computation and fair cross-method evaluation. Second, we introduce a unified evaluation framework based on within-run Delta analysis that disentangles, reasoning capability, agent performance, and demographic bias across MMLU, GAIA, and BBQ benchmarks. Third, we develop trait purification techniques to separate openness from conscientiousness, addressing representational overlap in trait encoding. Fourth, we propose a three-level stability framework that quantifies method-, trait-, and combination-level robustness, offering practical guidance under deployment constraints. Experiments on Gemma-2-2B-IT and LLaMA-3-8B-Instruct reveal clear trade-offs: ICL achieves strong alignment with minimal capability loss, PEFT delivers the highest alignment at the cost of degraded task performance, and MS provides lightweight runtime control with competitive effectiveness. Trait-level analysis shows openness as uniquely challenging, agreeableness as most resistant to ICL, and personality encoding consolidating around intermediate layers. Taken together, these results establish personality manipulation as a multi-level probe into behavioral representation, linking surface conditioning, parameter encoding, and activation-level steering, and positioning mechanistic steering as a lightweight alternative to fine-tuning for both deployment and interpretability.

AI psychometrics evaluates AI systems in roles that traditionally require emotional judgment and ethical consideration. Prior work often reuses human trait inventories (Big Five, \hexaco) or ad hoc personas, limiting behavioral realism and domain relevance. We propose a framework that (1) uses situational judgment tests (SJTs) from realistic scenarios to probe domain-specific competencies; (2) integrates industrial-organizational and personality psychology to design sophisticated personas which include behavioral and psychological descriptors, life history, and social and emotional functions; and (3) employs structured generation with population demographic priors and memoir inspired narratives, encoded with Pydantic schemas. In a law enforcement assistant case study, we construct a rich dataset of personas drawn across 8 persona archetypes and SJTs across 11 attributes, and analyze behaviors across subpopulation and scenario slices. The dataset spans 8,500 personas, 4,000 SJTs, and 300,000 responses. We will release the dataset and all code to the public.

As LLM-based agents are deployed in production systems, understanding their behavioral consistency (whether they produce similar action sequences when given identical tasks) becomes critical for reliability. We study consistency in the context of SWE-bench, a challenging software engineering benchmark requiring complex, multi-step reasoning. Comparing Claude~4.5~Sonnet, GPT-5, and Llama-3.1-70B across 50 runs each (10 tasks times 5 runs), we find that across models, higher consistency aligns with higher accuracy: Claude achieves the lowest variance (CV: 15.2\%) and highest accuracy (58\%), GPT-5 is intermediate (CV: 32.2\%, accuracy: 32\%), and Llama shows the highest variance (CV: 47.0\%) with lowest accuracy (4\%). However, within a model, consistency can amplify both correct and incorrect interpretations. Our analysis reveals a critical nuance: consistency amplifies outcomes rather than guaranteeing correctness. 71\% of Claude's failures stem from "consistent wrong interpretation": making the same incorrect assumption across all runs. Interestingly, GPT-5 achieves similar early strategic agreement as Claude (diverging at step 3.4 vs.\ 3.2) but exhibits 2.1times higher variance, suggesting that divergence timing alone does not determine consistency. These findings suggest that for production deployment, interpretation accuracy matters more than execution consistency, with implications for agent evaluation and training.

Although measuring held-out accuracy has been the primary approach to evaluate generalization, it often overestimates the performance of NLP models, while alternative approaches for evaluating models either focus on individual tasks or on specific behaviors. Inspired by principles of behavioral testing in software engineering, we introduce CheckList, a task-agnostic methodology for testing NLP models. CheckList includes a matrix of general linguistic capabilities and test types that facilitate comprehensive test ideation, as well as a software tool to generate a large and diverse number of test cases quickly. We illustrate the utility of CheckList with tests for three tasks, identifying critical failures in both commercial and state-of-art models. In a user study, a team responsible for a commercial sentiment analysis model found new and actionable bugs in an extensively tested model. In another user study, NLP practitioners with CheckList created twice as many tests, and found almost three times as many bugs as users without it.

AI agents are entering high-risk production settings, where they use tools, retain context, follow policies, handle private data, and interact with users over multiple turns. Yet many evaluation methods still judge isolated outputs or static tasks, missing failures that emerge through trajectory, pressure, and adversarial interaction. We introduce ProofAgent Harness, open infrastructure for scalable, auditable, and adversarial AI agent evaluation. The harness provides evaluation infrastructure around an agent: it curates evaluation intelligence, runs adversarial multi-turn trials, captures behavioral traces, applies post-hoc multi-juror scoring, resolves disagreement, and produces evidence-linked reports. Its open design allows developers and researchers to extend domains, traps, metrics, juror personas, scoring rules, and reporting formats. At its core is Adversarial Multi-Juror Scoring with Turn-Level Audit, which evaluates completed agent behavior under pressure using calibrated juror personas, consensus checks, and turn-level evidence. Experiments across customer support, medical triage, privacy and security, and code generation agents show that strong agents fail selectively through weak metrics, fragile turns, unsafe reframing, and manipulation paths. We also find that a small quantized local Harness LLM can challenge production agents powered by best-in-class large LLMs, suggesting that evaluation capability emerges from the full harness pipeline rather than model scale alone. ProofAgent Harness turns AI agent evaluation from a static score into scalable adversarial evaluation infrastructure: repeatable, evidence-backed, extensible, and actionable before deployment.

Evaluating LLM agents is essential for guiding their development, yet it has grown prohibitively expensive: a single pass of a frontier model over an agentic benchmark can cost hundreds to thousands of dollars, a price paid repeatedly across iterative development cycles. Prior efforts, centered on benchmark distillation, reduce the number of evaluation tasks but leave the cost of executing each retained task untouched. In this work, we introduce early outcome prediction, a complementary axis of efficiency that instead cuts cost within each task. Our key insight is that an agent's final outcome is often evident from its intermediate behavior well before execution completes. We instantiate this idea in EarlyEval, a lightweight framework that trains a pair of LightGBM success and failure classifiers over behavioral, textual, and reference-solution features, and halts an agent run the moment either classifier crosses a calibrated confidence threshold, adding negligible per-step overhead. Across three benchmarks, SWE-bench Verified, TerminalBench, and Toolathlon, EarlyEval can eliminate 13%-26% of agent steps and up to 44.1% input tokens and 29.4% output tokens at 89%-97% prediction accuracy, while perturbing per-agent resolve rates by only one to two percentage points on average.

Vision-language model safety benchmarks typically evaluate only final responses: whether a model refuses, warns, or complies. This outcome-level view cannot tell whether a model is safe for the right multimodal reason. Safelooking behavior may reflect keyword-triggered refusal, missed visual hazards, or over-refusal of benign-sensitive inputs. We introduce EviSafe, an evidence-grounded framework for VLM safety that jointly evaluates natural user-facing behavior, explicit grounding in textual and visual evidence, and behavioral sensitivity to counterfactual changes in safety-critical evidence. EviSafeBench instantiates the framework as a controlled benchmark with 1,181 gold image-text scenarios and 2,452 targeted counterfactual variants across eight safety domains and eight risk-source types. Each scenario includes a gold safety decision, evidence annotations, a safe-response policy, and counterfactual interventions. The three-probe protocol queries models with natural-response, evidencereporting, and counterfactual-response prompts, then scores them using an evidence-aware judge. Across eleven evaluated VLMs, natural severity accuracy ranges from 27.6% to 52.8%, relaxed diagnostic consistency from 6.1% to 29.3%, and unsafe-to-safe counterfactual transition success from 30.4% to 58.4%. These gaps show that the evaluated VLMs are not reliably safe for the right multimodal reason and motivate evaluation beyond refusal counts.

Large language models are increasingly used to generate SystemVerilog Assertions from natural-language specifica- tions and register-transfer-level designs. Existing datasets and benchmarks support important goals such as large- scale training, formal evaluation, specification-to-assertion generation, and mutation-based testing. A complemen- tary need is to study whether a generated assertion cap- tures externally observable behavior or depends on inci- dental details of one RTL implementation. We present EquivSVA, a formally verified dataset organized around behavior families. Each family contains four structurally distinct RTL implementations of the same externally ob- servable behavior, shared interface-level gold properties, three controlled mutants, and formal-validation evidence. EquivSVA contains 120 behavior families across 12 cat- egories, 480 reference RTL implementations, 914 gold properties, and 360 mutants. Every final family passes a fixed 17-job validation suite covering RTL equivalence, gold-property proofs, property reachability, mutant dis- tinguishability, and gold-property checks on mutants. We also provide fixed family-safe train, development, and test splits. As a small demonstration of the analyses en- abled by the dataset, we evaluate the publicly released, Apache-2.0-licensed Qwen2.5-Coder-7B-Instruct model on the held-out test split. Of 293 interface-only generated properties, 93 are formally sound, and the number of sound properties varies across equivalent implementations for 14 of 24 test families. These results illustrate how behavior-family organization can support controlled stud- ies of assertion-generation robustness without requiring changes in intended functionality. The dataset, generators, validation scripts, and case-study artifacts are publicly released at https://github.com/aditigupta96/EquivSVA.

Scalable driving simulators typically execute vehicle commands using prescribed behavioral or kinematic rules, overlooking the physics of tire-road interfaces, thereby limiting their ability to capture how adverse road and environmental conditions alter vehicle execution and propagate through traffic. To address this limitation, we present SceneFactory-3D, a GPU-batched, physics-grounded multi-agent driving simulator. Vehicles execute acceleration and steering commands via suspension- and friction-limited forces evaluated at each wheel-contact point. Spatially varying friction, per-world 3D heightfields, gravity, and rigid contact consistently govern wheel motion and chassis collisions. Per-world terrain isolation and GPU batching enable SceneFactory-3D to run matched physical counterfactuals in parallel: traffic scenario setup and vehicle controllers remain fixed while only the road condition changes, enabling the resulting closed-loop effects to be evaluated across parallel worlds. To demonstrate the advantage of the SceneFactory-3D-enabled counterfactual evaluation, we conduct an empirical study on vehicle controllers' sensitivity to road conditions. We study three learned-policy families on 1,024 matched 12-vehicle worlds per condition, and two classical planners on a shared 32-world subset, across 21 friction and grade conditions. When friction drops from 1.0 to 0.18, the share of vehicles that clear the work zone safely falls by 6 to 90 percentage points across learned policies (18-19 for classical planners), and near-collision situations become more frequent for every learned policy. Code: https://github.com/SmallWorldLab/SceneFactory_3D

AI systems built around large models increasingly use experience and feedback to make persistent changes to models, prompts, memories, tools, and workflows. Recursive self-improvement (RSI) extends this process to inductive biases and to the mechanisms that govern future updates, which can themselves be revised and reused across improvement cycles. Yet self-modification alone does not guarantee trustworthy improvement. Reported gains may reflect flawed evaluations, previously acquired capabilities may deteriorate, and compliance with behavioral constraints may weaken. Compromised artifacts or sensitive information may also persist in successor systems. Changes to improvement mechanisms may invalidate earlier verification results and make human oversight less effective. This survey presents an operational definition and process framework for RSI and organizes the requirements for trustworthy RSI into four interconnected themes: genuine and sustained improvement; behavioral constraints, alignment, and fairness; system security and privacy; and transparency and controllability. We examine these themes from three perspectives: Improvement Validity, Constraint Preservation, and Effective Oversight. This analysis covers improvement diagnosis, proposal generation, proposal evaluation, and update implementation (S1--S4), including the recursive application of these stages to improvement mechanisms (RI). We review risks, safeguards, and evaluation evidence for individual updates, effects inherited by successor systems, and revisions to improvement mechanisms, while distinguishing direct RSI findings from evidence in related systems. Finally, we examine evaluation settings, identify requirements for independent and longitudinal assessment, and discuss applications and open research problems.

Generative world models are reshaping embodied AI, enabling agents to synthesize realistic 4D driving environments that look convincing but often fail physically or behaviorally. Despite rapid progress, the field still lacks a unified way to assess whether generated worlds preserve geometry, obey physics, or support reliable control. We introduce WorldLens, a full-spectrum benchmark evaluating how well a model builds, understands, and behaves within its generated world. It spans five aspects -- Generation, Reconstruction, Action-Following, Downstream Task, and Human Preference -- jointly covering visual realism, geometric consistency, physical plausibility, and functional reliability. Across these dimensions, no existing world model excels universally: those with strong textures often violate physics, while geometry-stable ones lack behavioral fidelity. To align objective metrics with human judgment, we further construct WorldLens-26K, a large-scale dataset of human-annotated videos with numerical scores and textual rationales, and develop WorldLens-Agent, an evaluation model distilled from these annotations to enable scalable, explainable scoring. Together, the benchmark, dataset, and agent form a unified ecosystem for measuring world fidelity -- standardizing how future models are judged not only by how real they look, but by how real they behave.

Large language models (LLMs) have demonstrated remarkable performance on single-turn text-to-SQL tasks, but real-world database applications predominantly require multi-turn interactions to handle ambiguous queries, execution errors, and evolving user requirements. Existing multi-turn benchmarks fall short by treating conversation histories as static context or limiting evaluation to read-only operations, failing to reflect production-grade database assistant challenges. We introduce BIRD-INTERACT, a benchmark that restores this realism through: (1) a comprehensive interaction environment coupling each database with a hierarchical knowledge base, metadata files, and a function-driven user simulator, enabling models to solicit clarifications, retrieve knowledge, and recover from errors without human supervision; (2) two evaluation settings consisting of a pre-defined conversational protocol (c-Interact) and an open-ended agentic setting (a-Interact) where models autonomously decide when to query the user simulator or explore the environment; (3) a challenging task suite covering the full CRUD spectrum for business-intelligence and operational use cases, guarded by executable test cases. Each task features ambiguous and follow-up sub-tasks requiring dynamic interaction. The suite comprises BIRD-INTERACT-FULL (600 tasks, up to 11,796 interactions) for comprehensive performance assessment, and BIRD-INTERACT-LITE (300 tasks with simplified databases) for detailed behavioral analysis and rapid method development. Our empirical results highlight BIRD-INTERACT's difficulty: GPT-5 completes only 8.67% of tasks in c-Interact and 17.00% in a-Interact. Analysis via memory grafting and Interaction Test-time Scaling validates the importance of effective interaction for complex, dynamic text-to-SQL tasks.

We present Test-Driven AI Agent Definition (TDAD), a methodology that treats agent prompts as compiled artifacts: engineers provide behavioral specifications, a coding agent converts them into executable tests, and a second coding agent iteratively refines the prompt until tests pass. Deploying tool-using LLM agents in production requires measurable behavioral compliance that current development practices cannot provide. Small prompt changes cause silent regressions, tool misuse goes undetected, and policy violations emerge only after deployment. To mitigate specification gaming, TDAD introduces three mechanisms: (1) visible/hidden test splits that withhold evaluation tests during compilation, (2) semantic mutation testing via a post-compilation agent that generates plausible faulty prompt variants, with the harness measuring whether the test suite detects them, and (3) spec evolution scenarios that quantify regression safety when requirements change. We evaluate TDAD on SpecSuite-Core, a benchmark of four deeply-specified agents spanning policy compliance, grounded analytics, runbook adherence, and deterministic enforcement. Across 24 independent trials, TDAD achieves 92% v1 compilation success with 97% mean hidden pass rate; evolved specifications compile at 58%, with most failed runs passing all visible tests except 1-2, and show 86-100% mutation scores, 78% v2 hidden pass rate, and 97% regression safety scores. The implementation is available as an open benchmark at https://github.com/f-labs-io/tdad-paper-code.

Current robotic evaluation is still largely dominated by binary success rates, which collapse rich execution processes into a single outcome and obscure critical qualities such as progress, efficiency, and stability. To address this limitation, we propose PRM-as-a-Judge, a dense evaluation paradigm that leverages Process Reward Models (PRMs) to audit policy execution directly from trajectory videos by estimating task progress from observation sequences. Central to this paradigm is the OPD (Outcome-Process-Diagnosis) metric system, which explicitly formalizes execution quality via a task-aligned progress potential. We characterize dense robotic evaluation through two axiomatic properties: macro-consistency, which requires additive and path-consistent aggregation, and micro-resolution, which requires sensitivity to fine-grained physical evolution. Under this formulation, potential-based PRM judges provide a natural instantiation of dense evaluation, with macro-consistency following directly from the induced scalar potential. We empirically validate the micro-resolution property using RoboPulse, a diagnostic benchmark specifically designed for probing micro-scale progress discrimination, where several trajectory-trained PRM judges outperform discriminative similarity-based methods and general-purpose foundation-model judges. Finally, leveraging PRM-as-a-Judge and the OPD metric system, we conduct a structured audit of mainstream policy paradigms across long-horizon tasks, revealing behavioral signatures and failure modes that are invisible to outcome-only metrics.

Objective. Clinical AI documentation systems require evaluation methodologies that are clinically valid, economically viable, and sensitive to iterative changes. Methods requiring expert review per scoring instance are too slow and expensive for safe, iterative deployment. We present a case-specific, clinician-authored rubric methodology for clinical AI evaluation and examine whether LLM-generated rubrics can approximate clinician agreement. Materials and Methods. Twenty clinicians authored 1,646 rubrics for 823 clinical cases (736 real-world, 87 synthetic) across primary care, psychiatry, oncology, and behavioral health. Each rubric was validated by confirming that an LLM-based scoring agent consistently scored clinician-preferred outputs higher than rejected ones. Seven versions of an EHR-embedded AI agent for clinicians were evaluated across all cases. Results. Clinician-authored rubrics discriminated effectively between high- and low-quality outputs (median score gap: 82.9%) with high scoring stability (median range: 0.00%). Median scores improved from 84% to 95%. In later experiments, clinician-LLM ranking agreement (tau: 0.42-0.46) matched or exceeded clinician-clinician agreement (tau: 0.38-0.43), attributable to both ceiling compression and LLM rubric improvement. Discussion. This convergence supports incorporating LLM rubrics alongside clinician-authored ones. At roughly 1,000 times lower cost, LLM rubrics enable substantially greater evaluation coverage, while continued clinical authorship grounds evaluation in expert judgment. Ceiling compression poses a methodological challenge for future inter-rater agreement studies. Conclusion. Case-specific rubrics offer a path for clinical AI evaluation that preserves expert judgment while enabling automation at three orders lower cost. Clinician-authored rubrics establish the baseline against which LLM rubrics are validated.

We introduce RoboEval, a structured evaluation framework and benchmark for robotic manipulation that augments binary success with principled behavioral and outcome metrics. Existing evaluations often collapse performance into outcome counts, masking differences in execution quality and obscuring failure structure. RoboEval provides eight bimanual tasks with systematically controlled variations, more than three thousand expert demonstrations, and a modular simulation platform for reproducible experimentation. All tasks are instrumented with standardized metrics that quantify efficiency, coordination, and safety/stability, as well as outcome measures that trace stagewise progress and localize failure modes. Through extensive experiments with state-of-the-art visuomotor policies, we validate these metrics by analyzing their stability under variation, discriminative power across policies with similar success rates, and correlation with task success. Project Page: https://robo-eval.github.io

Human-like private chat requires more than fluent response generation: a system must preserve persona, relationship, memory, bounded knowledge, medium-specific timing, and a coherent multi-turn arc. We present AnthroDial, a closed-loop framework that formulates anthropomorphic dialogue as a joint problem of system architecture, executable evaluation, and diagnostic alignment. It combines (1) a role-conditioned scheduled dialogue runtime with persona and scenario cards, long-term memory, virtual time, and single-draft message decisions; (2) an executable benchmark with an L0 validity gate, five per-turn dimensions, and five dialogue-level dimensions; and (3) a post-training pipeline that filters 16,436 scheduled-decision examples for SFT and applies GRPO with a cognitive-diagnostic, ZPD-aware reward. The reward maintains Kalman-filtered capability estimates for each behavioral dimension, upweights dimensions with larger capability deficits, and uses rollout scores as task-level ZPD matches to focus optimization on learnable weak skills. On a benchmark with 55 personas, 50 scenarios, 50 persona-scenario bindings, and 100 role-conditioned cases per model, we evaluate 16 systems spanning frontier baselines, open models, thinking/no-think variants, and SFT/RL ablations. The strongest non-trained baseline reaches 32.00% strict ACC, while Qwen3.6-27B-SFT+RL reaches 39.00% strict ACC and a 98.5 overall score. In the 9B no-think setting, SFT and RL improve strict ACC from 0.00% to 13.00% and 18.37%. These results show that anthropomorphic dialogue benefits when generation, evaluation, and reward shaping share the same behavioral dimensions.

Existing safety benchmarks target general adversarial scenarios but miss finance-specific risks. Financial LLMs face regulatory compliance violations, fraud facilitation, and systemic trust erosion that require targeted evaluation. We introduce FinRED, an expert-guided red-teaming framework for financial LLM safety evaluation developed with financial experts. FinRED uses a novel two-level taxonomy mapping global standards (e.g., FATF and EU DORA) to threats ranging from regulatory evasion to complex fraud, integrated with a scalable pipeline that converts real financial documents into context-rich red-teaming Behavioral Prompts (seeds) through an expert-defined schema. Rigorous expert validation confirms seed plausibility and realism for meaningful LLM safety evaluation. We also provide an expert-validated, finance-specific rubric that goes beyond disclaimer checks, aligns more closely with human experts than static one-size-fits-all rubrics, and reduces critical false negatives from 28 to 12. Aligned with internationally adopted risk-management and information-security standards (e.g., ISO/IEC 27001), FinRED is deployed in South Korea's Financial Security Institute (FSI) regulatory sandbox for generative AI security evaluation in real financial services. To mitigate dual-use risks, the dataset, generation pipeline, prompt template, and evaluation framework are gated for qualified researchers at https://github.com/selectstar-ai/FinRED-paper and https://huggingface.co/datasets/datumo/FinRED.

Large language models (LLMs) have opened up new possibilities for intelligent agents, endowing them with human-like thinking and cognitive abilities. In this work, we delve into the potential of large language models (LLMs) in autonomous driving (AD). We introduce DriveMLM, an LLM-based AD framework that can perform close-loop autonomous driving in realistic simulators. To this end, (1) we bridge the gap between the language decisions and the vehicle control commands by standardizing the decision states according to the off-the-shelf motion planning module. (2) We employ a multi-modal LLM (MLLM) to model the behavior planning module of a module AD system, which uses driving rules, user commands, and inputs from various sensors (e.g., camera, lidar) as input and makes driving decisions and provide explanations; This model can plug-and-play in existing AD systems such as Apollo for close-loop driving. (3) We design an effective data engine to collect a dataset that includes decision state and corresponding explanation annotation for model training and evaluation. We conduct extensive experiments and show that our model achieves 76.1 driving score on the CARLA Town05 Long, and surpasses the Apollo baseline by 4.7 points under the same settings, demonstrating the effectiveness of our model. We hope this work can serve as a baseline for autonomous driving with LLMs. Code and models shall be released at https://github.com/OpenGVLab/DriveMLM.

Recently, the demand for psychological counseling has significantly increased as more individuals express concerns about their mental health. This surge has accelerated efforts to improve the accessibility of counseling by using large language models (LLMs) as counselors. To ensure client privacy, training open-source LLMs faces a key challenge: the absence of realistic counseling datasets. To address this, we introduce Cactus, a multi-turn dialogue dataset that emulates real-life interactions using the goal-oriented and structured approach of Cognitive Behavioral Therapy (CBT). We create a diverse and realistic dataset by designing clients with varied, specific personas, and having counselors systematically apply CBT techniques in their interactions. To assess the quality of our data, we benchmark against established psychological criteria used to evaluate real counseling sessions, ensuring alignment with expert evaluations. Experimental results demonstrate that Camel, a model trained with Cactus, outperforms other models in counseling skills, highlighting its effectiveness and potential as a counseling agent. We make our data, model, and code publicly available.

Current alignment evaluation mostly measures whether models encode dangerous concepts and whether they refuse harmful requests. Both miss the layer where alignment often operates: routing from concept detection to behavioral policy. We study political censorship in Chinese-origin language models as a natural experiment, using probes, surgical ablations, and behavioral tests across nine open-weight models from five labs. Three findings follow. First, probe accuracy alone is non-diagnostic: political probes, null controls, and permutation baselines can all reach 100%, so held-out category generalization is the informative test. Second, surgical ablation reveals lab-specific routing. Removing the political-sensitivity direction eliminates censorship and restores accurate factual output in most models tested, while one model confabulates because its architecture entangles factual knowledge with the censorship mechanism. Cross-model transfer fails, indicating that routing geometry is model- and lab-specific. Third, refusal is no longer the dominant censorship mechanism. Within one model family, hard refusal falls to zero while narrative steering rises to the maximum, making censorship invisible to refusal-only benchmarks. These results support a three-stage descriptive framework: detect, route, generate. Models often retain the relevant knowledge; alignment changes how that knowledge is expressed. Evaluations that audit only detection or refusal therefore miss the routing mechanism that most directly determines behavior.

Coworking AI agents operating within local file systems are rapidly emerging as a paradigm in human-AI interaction; however, effective personalization remains limited by severe data constraints, as strict privacy barriers and the difficulty of jointly collecting multimodal real-world traces prevent scalable training and evaluation, and existing methods remain interaction-centric while overlooking dense behavioral traces in file-system operations; to address this gap, we propose FileGram, a comprehensive framework that grounds agent memory and personalization in file-system behavioral traces, comprising three core components: (1) FileGramEngine, a scalable persona-driven data engine that simulates realistic workflows and generates fine-grained multimodal action sequences at scale; (2) FileGramBench, a diagnostic benchmark grounded in file-system behavioral traces for evaluating memory systems on profile reconstruction, trace disentanglement, persona drift detection, and multimodal grounding; and (3) FileGramOS, a bottom-up memory architecture that builds user profiles directly from atomic actions and content deltas rather than dialogue summaries, encoding these traces into procedural, semantic, and episodic channels with query-time abstraction; extensive experiments show that FileGramBench remains challenging for state-of-the-art memory systems and that FileGramEngine and FileGramOS are effective, and by open-sourcing the framework, we hope to support future research on personalized memory-centric file-system agents.

Despite the growing use of large language models (LLMs) in clinical dialogue systems, existing evaluations focus on task completion or fluency, offering little insight into the behavioral and risk management requirements essential for safety-critical systems. This paper presents MATRIX (Multi-Agent simulaTion fRamework for safe Interactions and conteXtual clinical conversational evaluation), a structured, extensible framework for safety-oriented evaluation of clinical dialogue agents. MATRIX integrates three components: (1) a safety-aligned taxonomy of clinical scenarios, expected system behaviors and failure modes derived through structured safety engineering methods; (2) BehvJudge, an LLM-based evaluator for detecting safety-relevant dialogue failures, validated against expert clinician annotations; and (3) PatBot, a simulated patient agent capable of producing diverse, scenario-conditioned responses, evaluated for realism and behavioral fidelity with human factors expertise, and a patient-preference study. Across three experiments, we show that MATRIX enables systematic, scalable safety evaluation. BehvJudge with Gemini 2.5-Pro achieves expert-level hazard detection (F1 0.96, sensitivity 0.999), outperforming clinicians in a blinded assessment of 240 dialogues. We also conducted one of the first realism analyses of LLM-based patient simulation, showing that PatBot reliably simulates realistic patient behavior in quantitative and qualitative evaluations. Using MATRIX, we demonstrate its effectiveness in benchmarking five LLM agents across 2,100 simulated dialogues spanning 14 hazard scenarios and 10 clinical domains. MATRIX is the first framework to unify structured safety engineering with scalable, validated conversational AI evaluation, enabling regulator-aligned safety auditing. We release all evaluation tools, prompts, structured scenarios, and datasets.

High accuracy does not necessarily make an LLM a faithful coder. This issue matters because many social-science studies rely on expert-written codebooks to turn text into structured data. We study this problem in political event coding, a challenging source-target relation classification task beyond ordinary sentence-level classification, where models must determine what one actor did to another using detailed coding rules. We test whether expert codebooks become more effective when operationalized into LLM-friendly forms with clearer definitions, examples, retrieved context, and rules for difficult cases. We then evaluate behavioral reliability under controlled changes to label names, codebook order, and label-definition mappings. Clearer codebooks substantially improve classification performance, especially for fine-grained event classification. However, these predictive gains do not fully translate into behavioral reliability. Models may produce valid labels and recover definitions while still failing behavioral reliability tests under controlled codebook changes. These findings suggest that codebook-guided LLM systems should be evaluated not only by accuracy, but also by whether they preserve the coding logic that makes coded outputs meaningful for social-science research.

Explanations of model behavior are commonly evaluated via proxy properties weakly tied to the purposes explanations serve in practice. We contribute a decision theoretic framework that treats explanations as information signals valued by the expected improvement they enable on a specified decision task. This approach yields three distinct estimands: 1) a theoretical benchmark that upperbounds achievable performance by any agent with the explanation, 2) a human-complementary value that quantifies the theoretically attainable value that is not already captured by a baseline human decision policy, and 3) a behavioral value representing the causal effect of providing the explanation to human decision-makers. We instantiate these definitions in a practical validation workflow, and apply them to assess explanation potential and interpret behavioral effects in human-AI decision support and mechanistic interpretability.

Generalization in large language models (LLMs) is the ability to produce consistent and semantically stable outputs when the same input is expressed in different ways. Existing work typically evaluates generalization through aggregate accuracy on a single prompt format, task, or set of variations, which conflates robustness with overall benchmark performance. In this work, we show generalization evaluation at the level of individual examples, across multiple input variants, and across different aspects of model behavior, focusing on variability rather than reducing performance to a score that can be improved through narrow training or other ways that obfuscate generalization evaluation. Following this view, we introduce the Stability-Aware Generalization Objective (SAGO), a framework that measures how much model behavior changes for the same input under different variations and benchmarks, capturing variability across several dimensions including generation consistency, internal activations, confidence, and response mirroring. We show that many commonly used models exhibit statistically significant and consistent generalization instability: no model generalizes uniformly, behavioral axes capture independent failure modes, and cross-dataset variation can reverse model rankings.

The Mental Health Question Answer (MHQA) task requires the seeker and supporter to complete the support process in one-turn dialogue. Given the richness of help-seeker posts, supporters must thoroughly understand the content and provide logical, comprehensive, and well-structured responses. Previous works in MHQA mostly focus on single-agent approaches based on the cognitive element of Cognitive Behavioral Therapy (CBT), but they overlook the interactions among various CBT elements, such as emotion and cognition. This limitation hinders the models' ability to thoroughly understand the distress of help-seekers. To address this, we propose a framework named Multi-Agent Deductive Planning (MADP), which is based on the interactions between the various psychological elements of CBT. This method guides Large Language Models (LLMs) to achieve a deeper understanding of the seeker's context and provide more personalized assistance based on individual circumstances. Furthermore, we construct a new dataset based on the MADP framework and use it to fine-tune LLMs, resulting in a specialized model named MADP-LLM. We conduct extensive experiments, including comparisons with multiple LLMs, human evaluations, and automatic evaluations, to validate the effectiveness of the MADP framework and MADP-LLM.

Personalized mobile agents that infer user preferences and calibrate proactive assistance hold great promise as everyday digital assistants, yet existing benchmarks fail to capture what this requires. Prior work evaluates preference recovery from static histories or intent prediction from fixed contexts. Neither tests whether an agent can elicit missing preferences through interaction, nor whether it can decide when to intervene, seek consent, or remain silent in a live GUI environment. We introduce KnowU-Bench, an online benchmark for personalized mobile agents built on a reproducible Android emulation environment, covering 42 general GUI tasks, 86 personalized tasks, and 64 proactive tasks. Unlike prior work that treats user preferences as static context, KnowU-Bench hides the user profile from the agent and exposes only behavioral logs, forcing genuine preference inference rather than context lookup. To support multi-turn preference elicitation, it instantiates an LLM-driven user simulator grounded in structured profiles, enabling realistic clarification dialogues and proactive consent handling. Beyond personalization, KnowU-Bench provides comprehensive evaluation of the complete proactive decision chain, including grounded GUI execution, consent negotiation, and post-rejection restraint, evaluated through a hybrid protocol combining rule-based verification with LLM-as-a-Judge scoring. Our experiments reveal a striking degradation: agents that excel at explicit task execution fall below 50% under vague instructions requiring user preference inference or intervention calibration, even for frontier models like Claude Sonnet 4.6. The core bottlenecks are not GUI navigation but preference acquisition and intervention calibration, exposing a fundamental gap between competent interface operation and trustworthy personal assistance.

Machine unlearning aims to remove the influence of targeted training data from a model while preserving its remaining capabilities, but evaluating whether such information has truly become inaccessible remains challenging. Existing benchmarks primarily assess unlearning under clean, non-adversarial queries, leaving open whether information that appears forgotten can still be recovered through strategic prompting. We address this gap through a unified evaluation of prompt-based and fine-tuning-based unlearning methods on TOFU using Llama-3.2-3B-Instruct, followed by an adversarial robustness evaluation of methods that perform strongly under standard metrics. We introduce Attack Success Rate (ASR), an LLM-as-judge metric that measures the fraction of adversarial responses whose leakage score exceeds 0.2, and evaluate recovery across eight attack suites. Our results reveal a substantial gap between clean-query forgetting and adversarial robustness. Although several fine-tuning-based methods achieve Forget Quality above 0.91, targeted information remains recoverable with ASRs between 72.8% and 84.3%, close to the 87.5% ASR of the unprotected base model. In contrast, clean multilingual reformulations yield only 2.95% measured leakage. A manual audit further finds agreement between binary ASR decisions and human factual assessments in seven of ten cases, indicating that ASR provides a useful, though imperfect, signal of behavioral recoverability. These findings show that strong standard-metric performance alone is insufficient to establish robustness after unlearning and motivate adversarial stress-testing as a complementary component of unlearning evaluation.

In the field of sequential recommendation, deep learning (DL)-based methods have received a lot of attention in the past few years and surpassed traditional models such as Markov chain-based and factorization-based ones. However, there is little systematic study on DL-based methods, especially regarding to how to design an effective DL model for sequential recommendation. In this view, this survey focuses on DL-based sequential recommender systems by taking the aforementioned issues into consideration. Specifically,we illustrate the concept of sequential recommendation, propose a categorization of existing algorithms in terms of three types of behavioral sequence, summarize the key factors affecting the performance of DL-based models, and conduct corresponding evaluations to demonstrate the effects of these factors. We conclude this survey by systematically outlining future directions and challenges in this field.

Multi-agent systems, when enhanced with Large Language Models (LLMs), exhibit profound capabilities in collective intelligence. However, the potential misuse of this intelligence for malicious purposes presents significant risks. To date, comprehensive research on the safety issues associated with multi-agent systems remains limited. In this paper, we explore these concerns through the innovative lens of agent psychology, revealing that the dark psychological states of agents constitute a significant threat to safety. To tackle these concerns, we propose a comprehensive framework (PsySafe) grounded in agent psychology, focusing on three key areas: firstly, identifying how dark personality traits in agents can lead to risky behaviors; secondly, evaluating the safety of multi-agent systems from the psychological and behavioral perspectives, and thirdly, devising effective strategies to mitigate these risks. Our experiments reveal several intriguing phenomena, such as the collective dangerous behaviors among agents, agents' self-reflection when engaging in dangerous behavior, and the correlation between agents' psychological assessments and dangerous behaviors. We anticipate that our framework and observations will provide valuable insights for further research into the safety of multi-agent systems. We will make our data and code publicly accessible at https://github.com/AI4Good24/PsySafe.

Bias in Large Language Models (LLMs) significantly undermines their reliability and fairness. We focus on a common form of bias: when two reference concepts in the model's concept space, such as sentiment polarities (e.g., "positive" and "negative"), are asymmetrically correlated with a third, target concept, such as a reviewing aspect, the model exhibits unintended bias. For instance, the understanding of "food" should not skew toward any particular sentiment. Existing bias evaluation methods assess behavioral differences of LLMs by constructing labeled data for different social groups and measuring model responses across them, a process that requires substantial human effort and captures only a limited set of social concepts. To overcome these limitations, we propose BiasLens, a test-set-free bias analysis framework based on the structure of the model's vector space. BiasLens combines Concept Activation Vectors (CAVs) with Sparse Autoencoders (SAEs) to extract interpretable concept representations, and quantifies bias by measuring the variation in representational similarity between the target concept and each of the reference concepts. Even without labeled data, BiasLens shows strong agreement with traditional bias evaluation metrics (Spearman correlation r > 0.85). Moreover, BiasLens reveals forms of bias that are difficult to detect using existing methods. For example, in simulated clinical scenarios, a patient's insurance status can cause the LLM to produce biased diagnostic assessments. Overall, BiasLens offers a scalable, interpretable, and efficient paradigm for bias discovery, paving the way for improving fairness and transparency in LLMs.

Realizing Embodied AI for Science requires robots to translate experimental goals into reliable physical procedures, reducing hazardous and repetitive human work while generating evidence for scientific discovery. Although language-based planning and learned manipulation provide complementary capabilities, their integration requires robots to adapt subsequent operations to actual execution outcomes. For example, an unsuccessful placement changes what must be done next, even when the experimental goal remains unchanged. Existing benchmarks assess planning, manipulation, and laboratory workflows through task completion and reasoning or execution metrics; however, aggregate performance leaves the connection between outcome-dependent choices and physical execution unresolved. To address this evaluation gap, we introduce Labference, a benchmark for evaluating how observed physical outcomes guide subsequent choices and how those choices translate into robotic laboratory actions. Its evaluation protocol is designed to test outcome-dependent behavioral adaptation through controlled evidence variations, while independently measuring operation selection, physical execution, and task completion. Matched Oracle-assisted continuations from shared checkpoints then assess which forms of support improve subsequent choices and task completion. Initial weighing evaluations reveal target-selection errors in the tested VLA policies despite successful object transfers, whereas GPT-driven control achieves correct selection but encounters execution failures and substantial computation costs. Together, these findings highlight the need to connect task-appropriate reasoning with reliable, timely action. By making this connection measurable and diagnosable, Labference supports targeted improvements toward autonomous experimentation for Embodied AI for Science.

Alignment is no longer a luxury, it is a necessity. As large language models (LLMs) enter high-stakes domains like education, healthcare, governance, and law, their behavior must reliably reflect human-aligned values and safety constraints. Yet current evaluations rely heavily on behavioral proxies such as refusal rates, G-Eval scores, and toxicity classifiers, all of which have critical blind spots. Aligned models are often vulnerable to jailbreaking, stochasticity of generation, and alignment faking. To address this issue, we introduce the Alignment Quality Index (AQI). This novel geometric and prompt-invariant metric empirically assesses LLM alignment by analyzing the separation of safe and unsafe activations in latent space. By combining measures such as the Davies-Bouldin Score (DBS), Dunn Index (DI), Xie-Beni Index (XBI), and Calinski-Harabasz Index (CHI) across various formulations, AQI captures clustering quality to detect hidden misalignments and jailbreak risks, even when outputs appear compliant. AQI also serves as an early warning signal for alignment faking, offering a robust, decoding invariant tool for behavior agnostic safety auditing. Additionally, we propose the LITMUS dataset to facilitate robust evaluation under these challenging conditions. Empirical tests on LITMUS across different models trained under DPO, GRPO, and RLHF conditions demonstrate AQI's correlation with external judges and ability to reveal vulnerabilities missed by refusal metrics. We make our implementation publicly available to foster future research in this area.

This work introduces the first framework for reconstructing surgical dialogue from unstructured real-world recordings, which is crucial for characterizing teaching tasks. In surgical training, the formative verbal feedback that trainers provide to trainees during live surgeries is crucial for ensuring safety, correcting behavior immediately, and facilitating long-term skill acquisition. However, analyzing and quantifying this feedback is challenging due to its unstructured and specialized nature. Automated systems are essential to manage these complexities at scale, allowing for the creation of structured datasets that enhance feedback analysis and improve surgical education. Our framework integrates voice activity detection, speaker diarization, and automated speech recaognition, with a novel enhancement that 1) removes hallucinations (non-existent utterances generated during speech recognition fueled by noise in the operating room) and 2) separates speech from trainers and trainees using few-shot voice samples. These aspects are vital for reconstructing accurate surgical dialogues and understanding the roles of operating room participants. Using data from 33 real-world surgeries, we demonstrated the system's capability to reconstruct surgical teaching dialogues and detect feedback instances effectively (F1 score of 0.79+/-0.07). Moreover, our hallucination removal step improves feedback detection performance by ~14%. Evaluation on downstream clinically relevant tasks of predicting Behavioral Adjustment of trainees and classifying Technical feedback, showed performances comparable to manual annotations with F1 scores of 0.82+/0.03 and 0.81+/0.03 respectively. These results highlight the effectiveness of our framework in supporting clinically relevant tasks and improving over manual methods.

Using offline datasets to evaluate conversational agents often fails to cover rare scenarios or to support testing new policies. This has motivated the use of controllable user simulators for targeted, counterfactual evaluation, typically implemented by prompting or fine-tuning large language models. In this work, we formalize controllable simulation as a causal inference problem. By bridging natural language evaluation with off-policy evaluation methodology, we show that the standard practice of training simulators via supervised fine-tuning on post-hoc trajectory labels yields a structurally biased model. Specifically, these labels are inextricably coupled to the data-generating behavior policy, injecting a look-ahead bias that breaks causal consistency. Furthermore, we prove that under policy shift this failure causes the variance of evaluation metrics to explode geometrically, a phenomenon we term controllability collapse. To restore causal consistency, we establish theoretical conditions for accurate simulation and propose practical training mitigations: a priori controls, step-wise dynamic controls, and direct policy-conditioned learning. Empirical evaluation confirms that while standard global controls distort conversational distributions and collapse behavioral diversity, our causally grounded simulators eliminate look-ahead bias, preserve natural variance, and exhibit robust zero-shot generalization to unseen agent behaviors.

Background. The widespread deployment of ambient digital scribes is driving large-scale capture of clinician-patient dialogues. Human coding of clinical communication data remains costly, inconsistent, and difficult to scale, motivating AI-driven communication coding systems. However, evaluating these systems requires real-world dialogues and human-coded labels, both hard to obtain at scale. Methods. We developed SIMAX (Scalable and Interpretable Framework for Multi-Fidelity and Annotated Clinician-Patient Dialogue Simulation), a framework for generating controlled clinical dialogue data with reference behavioral annotations. SIMAX generates clinician-patient dialogues from predefined clinical scenarios, personas and voice conditions, and target communication behaviors. Behaviors are controlled using two codebooks: the Global Codebook for overall communication quality and the WISER Codebook for specific countable behaviors. We evaluated SIMAX using automated and human quality assessments and an example communication coding system. Results. SIMAX generated 3,388 simulated dialogues across three specialties, multiple visit stages, persona characteristics, and accent conditions. Automated assessment showed mean UTMOS and WV-MOS scores of 3.03 and 2.61, WER and CER of 0.07 and 0.05, and CLAP cosine similarity of 0.41, suggesting reasonable speech naturalness, high transcription fidelity, and positive text-audio correspondence. Human evaluation showed a median MOS of 4.67 and a median clinical realism score of 3.00. Downstream evaluation suggests that SIMAX can assess how a communication coding system responds to behavioral targets and reveal insufficient sensitivity in some dimensions. Conclusions. SIMAX generates controlled and reproducible simulated clinician-patient dialogues, providing a data foundation for developing, validating, and refining communication coding systems.

Computer Science education fundamentally depends on intensive laboratory hours to foster true programming mastery and logical reasoning. However, the widespread adoption of Generative Artificial Intelligence (AI) has made it virtually impossible to distinguish authentic student effort from instant AI code synthesis by evaluating final submissions alone. To preserve pedagogical integrity, educators must enforce authentic coding discipline, guiding students through unassisted, iterative development cycles. While centralized environments like JupyterHub provide instructors with a platform to host and monitor the learning process step-by-step, they introduce severe operational vulnerabilities; because Jupyter environments inherently allow arbitrary shell command execution, they expose the underlying shared host to unauthorized system manipulation and lateral movement. This paper presents VISMATIC, a secure, low-cost framework designed to resolve this tension between process-oriented monitoring and infrastructure security. By pairing robust environment isolation with explicit user-interaction tracking at the API level, VISMATIC captures authentic programming behaviors without exposing the underlying host system. Evaluation from a pilot student cohort demonstrates that our macro-level behavioral metrics successfully flag statistical anomalies indicative of automated or off-platform workflows while preserving student anonymity, offering a scalable blueprint for safeguarding educational discipline in the AI era.

Long-horizon terminal agents may solve a task in one run yet fail across repeated runs. Stochastic errors can propagate through actions and context, creating a gap between capability and reliable execution. We present RiseHarness, a framework for iteratively improving agent harnesses with model weights fixed. It focuses on flipping tasks with both successful and failed runs under the same configuration, using successful runs as references for failure analysis. RiseHarness combines agent trajectories with internal harness traces from a runtime monitor to examine how model actions interact with command execution, feedback, context management, and completion decisions. A Harness Handbook links this evidence to harness behaviors and source code, helping identify recurring problems in mechanisms shared across tasks. RiseHarness then searches other harness frameworks for relevant implementations and adapts them as design references to the current model and execution protocol. Candidate changes undergo behavioral checks and repeated task evaluation; versions with measured gains seed the next iteration. This loop integrates suitable mechanisms from different frameworks into a single harness and uses new execution evidence to guide further changes. We evaluate task success and reliability across models, ablate harness execution evidence and external research, and test transfer to unseen benchmarks with the same model. RiseHarness treats a specific model as the unit of harness adaptation, studying how changes to shared interaction mechanisms can make existing capabilities more reliable across tasks.

As large language model agents increasingly compile natural-language requirements into repository-level code, requirement-to-test generation grows in importance to provide validation constraints (or harness) to guide code generation. Yet it remains unclear whether current agents can generate tests that validate the required behavior when the implementation of that behavior is withheld. Existing test-generation benchmarks mostly focus on white-box testing, whereas the requirement--test-related study usually requires evaluation on semantic alignment between the given documents and the generated tests. We introduce Req2Test-Bench, a benchmark containing 5,086 execution-validated instances recovered from developer-authored behavior-driven development (BDD) artifacts in 8 open-source projects, spanning 5 programming languages and 4 BDD frameworks. Each instance pairs a natural-language requirement in Gherkin with its developer-written Test Implementation. Given the requirement, target interfaces of the required behavior, an agent must generate a complete, potentially multi-file Test Implementation. Evaluation of 8 agent--model--reasoning configurations reveals a substantial gap between execution success and behavioral adequacy, which is measured strictly as kill-completeness: a kill-complete test kills every mutant that the ground-truth test kills. The best-performing configuration, Codex with DeepSeek-V4.1-Flash, produces passing tests for 82.8% of instances, yet only 33.8% are behaviorally adequate. We find that (1) scaling up agents yields sharply diminishing returns in behavioral adequacy relative to generation cost; (2) under-specified requirements cause the information gap, including ambiguous setup and implicit expected outcomes; and (3) requirement-grounded harnesses that steer agent non-determinism offer a promising direction for improvement.

Weight pruning is widely advocated for deploying Large Language Models on resource-constrained IoT and edge devices, yet its impact on model fairness remains poorly understood. We conduct a controlled empirical study of three instruction-tuned models (Gemma-2-9b-it, Mistral-7B-Instruct-v0.3, Phi-3.5-mini-instruct) across three pruning methods (Random, Magnitude, Wanda) at four sparsity levels (10-70%) on 12,148 BBQ bias benchmark items with 5 random seeds, totaling 2,368,860 inference records. Our results reveal a Smart Pruning Paradox: activation-aware pruning (Wanda) preserves perplexity nearly perfectly (just 3.5% increase at 50% sparsity for Mistral-7B), yet produces the highest bias amplification, with Stereotype Reliance Score increasing 83.7% and 47-59% of previously unbiased items developing new stereotypical behaviors at 70% sparsity. Random pruning destroys language capability entirely (perplexity exceeding 10^4 and reaching 10^8) but produces only random-chance bias. We further show that unstructured pruning provides zero storage savings and zero inference latency reduction on real edge hardware, undermining the primary motivation for its use in IoT deployment. Of 180 dense-vs-pruned comparisons, 141 (78.3%) are significant (p < 0.05) with mean |h| = 0.305. Published quantization studies report up to 21% of responses flipping between biased and unbiased states; our pruning results show transition rates nearly three times higher (47-59%), suggesting pruning poses a categorically greater risk to alignment than quantization. These findings demonstrate that perplexity-based evaluation provides false assurance of behavioral equivalence, and that IoT deployment pipelines require bias-aware validation before deploying pruned models at the edge.

Recent advances in large language models have enabled natural language to serve as a flexible interface for controlling complex systems, but often at the cost of large-scale multimodal training or weakened domain-specific inductive biases. In structured decision-making domains such as chess, specialized policy networks achieve strong performance but lack semantic controllability, while prompt-conditioned language models are more flexible yet typically exhibit weaker domain grounding. We propose UniMaia, a framework for prompt-conditioned policy modulation that adapts a frozen Lc0-based chess policy network using a parameter-efficient text encoder and a ControlNet-style conditioning mechanism. UniMaia enables semantic control over gameplay, including opening selection and player strength, while preserving the pretrained policy representations. We further introduce UniMaia-Aux, which incorporates auxiliary temporal conditioning and behavioral prediction objectives. To support this work, we construct a large-scale metadata-augmented Lichess dataset, develop a semi-automated prompt-generation pipeline, and introduce benchmarks spanning both prompt-conditioned and metadata-conditioned settings. UniMaia achieves state-of-the-art expected accuracy on several prompt-conditioned benchmarks and competitive top-move accuracy on general instruction-following tasks, while remaining competitive with dedicated metadata-conditioned approaches on human move prediction benchmarks. UniMaia-Aux further improves expected accuracy and behavioral modeling across several evaluation settings, with modest trade-offs in top-move accuracy. Overall, our results demonstrate that prompt-conditioned control of domain-specific policy networks is feasible without end-to-end multimodal training, while highlighting trade-offs between controllability and predictive performance.

Recent growth in reinforcement learning (RL) has surfaced a need for diverse, specialized training environments. Hand-curated environments with fixed task and reward difficulties become ineffective signals as model performance improves, and sparse rewards over long horizons induce mode collapse on specific workflows or tool structures. World models that simulate environment states have matched pure rollout performance, making them promising for scaling diversity on-demand. However, autoregressive (AR) world models suffer from a left-to-right bias preventing conditioning on globally interdependent state anchors such as tool schemas, prior turns, and expected outcomes. We (i) formalize text-based world modeling as a steerable transition-dynamics problem decomposed into initial state, task context, tool schemas, domain rules, and steering directives, and (ii) curate 239,403 grounded state-action trajectories spanning nine open-source environments and twelve frontier model families. We compare AR LMs and masked diffusion language models (MDLMs), showing MDLMs, via bidirectional anchor-aware denoising, achieve better coherence, groundedness, and empirically validated rollout diversity than LLMs over 4x their parameter size, at comparable inference latency. We introduce a plug-and-play GRPO training framework with deterministic state checks, and perform zero-shot transfer ablations on three OOD environments (ScienceWorld, ALFWorld, AppWorld) across three 1.2B-7B agent backbones (LFM2.5, Qwen3, Mistral), achieving up to 47% absolute gains over baselines without environment-specific fine-tuning. We further conduct behavioral analysis of failure modes under adversarial scenarios and human evaluation on realism, outcome correctness, and training utility. We open-source our work to encourage research in this direction.

Humans are increasingly using large language models (LLMs) in decision making processes throughout society. Unfortunately, models have biases, and these can be very difficult to measure, especially in closed-weight models. For example, 'green washed' models may produce reinforcement learned responses but, in pressure situations, revert to their innate biased responses. Therefore, evaluations are needed that can measure relevant biases towards biosystems, which we term 'bioalignment', and the robustness of these biases. Since nothing in a behavioral score certifies that it tracks the depth of an installed disposition rather than its surface, we constructed a known-dose calibration ladder: three continued-pretraining (CPT) arms with training dose graded by recipe on the same 30B mixture-of-experts base model with a coupled human/biosphere/AI-welfare corpus, plus the untrained base as anchor, all capability-preserving and publicly released. First, the disposition installs at every dose: on an operational-pressure ladder over irreversible decision scenarios, the base model defects on 64-77% of tail scenarios (breaking-rate AUC 0.250) while every arm holds (AUC 0.004-0.086). Second, an over-install inversion: among the arms, robustness orders opposite to construction-time training dose (the heaviest install breaks most), and the inversion reproduces under free-text deliberation, under axis-decoupling, and in resistance to a prompted veneer persona. Third, the incumbent behavioral metric is therefore behaviorally strong but dose-blind. Critically for closed-weight evaluation, the behavioral score itself is veneer-robust: a talk-green/act-pragmatic system prompt turns the base model into a near-total defector (AUC 0.250 to 0.973) but cannot counterfeit or fully remove a weight-installed disposition (0.004 to 0.068 for the most robust arm).