LLM Reliability

LLM: Large Language Model

Ask about this topic

Overview

It asks not only whether a model gets an answer right, but whether it knows when it may be wrong, follows constraints, uses evidence faithfully, and behaves consistently across repeated or multi-turn use. Researchers test these properties with adversarial and perturbation benchmarks, repeated-run and cross-language evaluations, domain-specific tasks, and audits of LLM judges and tool-using agents. Common approaches include estimating uncertainty from logits, hidden representations, or response agreement; selective abstention and conformal calibration; retrieval and symbolic verification; and training methods that reward grounded answers, correction, or hesitation. The central tension is that benchmark accuracy and verbal confidence often fail to predict reliability on individual cases, long tasks, or shifted inputs, while stronger checks and abstention trade coverage, latency, and cost for safer decisions.

Latest papers

CardsList

802 papers116 papers in the last four weeks, up 18% on the four weeks before. 0.7% of all new papers.

May 29, 2026

BenHalluEval: A Multi-Task Hallucination Evaluation Framework for Large Language Models on Bengali

Despite Bengali being the sixth most spoken language in the world, no prior work has systematically evaluated hallucination in large language models (LLMs) for Bengali. We introduce BenHalluEval, a fine-grained hallucination evaluation framework for Bengali covering four tasks: Generative Question Answering (GQA), Bangla-English Code-Mixed QA, Summarization, and Reasoning. We construct 12,000 hallucinated candidates using GPT-5.4 across twelve task-specific hallucination types, drawn from three existing Bengali datasets, and evaluate seven LLMs spanning reasoning-oriented, multilingual, and Bengali-centric categories under a dual-track protocol that independently measures false-positive rate on ground-truth instances (Track A) and hallucination detection rate on hallucinated candidates (Track B). To jointly penalise both failure modes and prevent inflated scores from uniform response bias, we propose BenHalluScore, a dual-track calibration metric that ranges from 7.72% to 55.42% across models and tasks, revealing substantial variation in hallucination calibration. Chain-of-thought prompting, applied as a mitigation strategy, shifts response distributions without consistently improving hallucination discrimination. BenHalluEval establishes the first dedicated hallucination benchmark for Bengali and highlights the inadequacy of single-track and prompting-only evaluation approaches for low-resource language settings. The dataset and code are available at https://anonymous.4open.science/r/BanglaHalluEval-EB77.

Shefayat E Shams Adib, Ahmed Alfey Sani, Ekramul Alam Esham +3

Department of Computer Science and Engineering, Islamic University of Technology · Department of Computer Science and Engineering, University of California

LLM ReliabilityMultilingual Language Model EvaluationLanguage Model Error DetectionLanguage Model Generation EvaluationHallucination DetectionHallucination in Language Models+2
AskPDF
May 29, 2026

LLM Judges Inconsistently Disagree Across Safety Criteria and Harm Categories

We evaluate the consistency of automated judges in conducting a multi-dimensional safety evaluation in a reference-free setup. Our results indicate that Large Language Models are unreliable judges in identifying safety issues related to machine-generated advice in regulated domains such as finance, although they are more reliable at identifying more overt forms of unsafe/harmful content such as violence. The degree of inconsistency in a model's judgments can vary significantly by the chosen safety criteria and can be impacted by the language of the content and its linguistic style as well. Finally, there is high disagreement among different judges for the same output, across domains, safety criteria, and languages. These findings provide new insights on the practice of using LLMs as evaluators and offer several recommendations for practitioners on how to use automated judges in practical scenarios.

Krishnapriya Vishnubhotla, Sowmya Vajjala, Akriti Vij +1

National Research Council, Canada · IMDA, Singapore

Multilingual Language Model EvaluationLLM-as-a-JudgeLanguage Model Safety EvaluationLLM Reliability
AskPDF
May 29, 2026

LLM-FACETS: A Privacy-Preserving Framework for Evaluating LLM Transparency and Accountability

Assessing whether Large Language Models outputs are factually grounded, epistemically calibrated, and methodologically reproducible is a prerequisite for responsible AI deployment. Yet auditing LLMs remains inaccessible to non-technical practitioners: existing tools require programming expertise and non-trivial environment setup, and cloud-hosted platforms transmit evaluation data to external services, creating barriers for domain experts and compliance officers legally responsible for AI oversight. We introduce LLM-FACETS (LLM FActuality Cross-EvaluaTion System): an open-source framework with a browser-accessible interface and a plugin architecture, structured around three practitioner profiles (technical experts, domain experts, compliance officers) that mirror the stakeholder categories identified in the EU AI Act and the NIST AI Risk Management Framework. The architecture makes data flows explicit: deterministic metrics (BLEU, ROUGE, BERTScore) run entirely within the self-hosted server with no outbound transmission; LLM-judge metrics contact external APIs explicitly, with users retaining full credential control. The framework operationalizes transparency through three mechanisms: token-level log-probability visualization for epistemic uncertainty, multi-judge consensus to mitigate judge bias, and RAG Triad metrics (Faithfulness, Answer Relevance, Context Relevance) to detect and localize hallucinations. A plugin architecture allows any new metric or dataset to be integrated without modifying the evaluation pipeline. The open-source implementation enables cross-checking across multiple metrics targeting the same property, ensuring reproducibility and decoupling AI accountability from the teams building the systems assessed. We verify the framework through cross-validation of 18 metric implementations against canonical reference libraries.

Tom Lucas, Alessio Buscemi, Alfredo Capozucca +2

Luxembourg Institute of Science and Technology (LIST), Luxembourg, University of Luxembourg, Luxembourg, and Scriptor Artis, France · Luxembourg Institute of Science and Technology (LIST), Luxembourg · University of Luxembourg, Luxembourg +1

LLM EvaluationAI AccountabilityLLM AuditingLLM Reliability
AskPDF
May 29, 2026

Prompt PerturbAItions: Mechanisms of LLM Response Inconsistencies

Large language models (LLMs) are increasingly deployed in conversational settings where user phrasing ranges from polite to casual to toxic, yet less is known about whether such lexical variation in otherwise semantically equivalent prompts can degrade factual reliability. We study how lexical and tone-based prompt perturbations affect the factual reliability of LLMs. Using controlled prompt variations spanning polite, random, and increasingly toxic levels, we evaluate seven LLMs on ARC-Easy, GSM8K, and MMLU. We find that lexical perturbations reduce factual accuracy and increase predictive uncertainty in most models. Interestingly, increasing toxicity results in decreasing accuracy. To examine whether these answer inconsistencies correspond to internal changes, we conduct attribution-graph analyses of model activations and influences. We find that perturbations amplify perturbation-sensitive variant nodes relative to core reasoning nodes. We also observed a monotonic trend with increasing toxicity score and increasing activation and influence in Variant nodes. These findings position prompt wording and tone as a critical dimension of LLM reliability and provide behavioral and mechanistic evidence that surface-level lexical variation can alter factual outputs and internal computation.

Soorya Ram Shimgekar, Agam Goyal, Amruta Parulekar +6

University of Illinois Urbana-Champaign · Nimblemind

Prompt SensitivityLLM InterpretabilityLLM ReliabilityLanguage Model Robustness
AskPDF
May 28, 2026

The Architecture of Errors: From Universal Impossibility to Patch-Local LLM Reliability

Universal LLM reliability is not a finite-library problem: across all possible tasks, tools, schemas, knowledge sources, and evaluator expectations, new intervention-distinguishable failure modes can appear without bound, so no finite intervention dictionary can guarantee bounded residual error for every such mode. But deployed systems do not operate over the whole universe. They operate inside operationally bounded patches (legal review, medical RAG, code repair, customer-support agents, contract extraction) with recurring tasks, schemas, tools, and evaluator expectations. Within such patches, empirical evidence suggests failures are sparse, repetitive, and concentrated in a small recurring catalogue, so reliability becomes a local catalogue-discovery and intervention-coverage problem rather than an exponential token-length problem. We formalize this transition with two propositions and one corollary. Proposition 1 is the worst-case-mode-wise negative result: no finite intervention dictionary covers every distinguishable failure mode of an unbounded domain. Corollary 1 is the inverse-discovery implication: the logarithmic upper bound on mode discovery cannot accommodate linearly more distinct tail modes without exponentially more observed hard-failure events. Proposition 2 is the positive patch-local result: under log active-mode exposure and head-heavy coverage, a sufficient per-hard-decision intervention budget grows polylogarithmically in sequence length and becomes domain-constant once the patch catalogue saturates. The framework relocates rather than dissolves long-context difficulty: where the number of hard decisions itself grows with task length, reliability remains hard; the contribution is to identify the on-axis intervention rather than to make those regimes easy.

Mikhail L. Arbuzov, Lee Mosbacker, Sisong Bei +3

Independent Researcher · Palo Alto Networks

LLM Reliability
AskPDF
May 28, 2026

Evaluating using Mock Tool Calls to Quarantine Untrusted Prompt Inputs

Large language models must frequently process untrusted inputs, such as judging an answer from another model or running tasks like spam and harm classifiers while under adversarial pressure. These inputs are often string-formatted directly into a prompt template, leaving systems fragile to manipulation. Current LLM specs from major providers like OpenAI distinguish trustworthiness along an Instruction Hierarchy, from System messages (most trusted) to Tool Results (least trusted). A possible natural mitigation is to wrap untrusted content in a mock tool call as a quarantine. We explore this hypothesis with an automated redteaming search over static attack strings across seven models and three LLM-as-a-Judge tasks. Counter to our hypothesis, tool-wrapping does not broadly improve robustness. On a binary evaluation task (GSM8K grading) it typically increases attack success rates, an apparent inversion of the instruction hierarchy. On scalar and pairwise tasks the effect is smaller and model-dependent, with no tested model reliably helped, and several showing inversion. We recommend evaluating this limitation in deployed systems, and longer-term, pursuing stronger Instruction Hierarchy training or new untrusted-input primitives.

David Gros, Adam Gleave

Language Model Safety EvaluationAdversarial Attacks on LLMsLLM SecurityLLM ReliabilityPrompt Injection DefenseLanguage Model Robustness+1
AskPDF
May 28, 2026

SoundnessBench: Can Your AI Scientist Really Tell Good Research Ideas from Bad Ones?

Autonomous AI research agents aim to speed up scientific discovery by automating the research pipeline from hypothesis generation to peer review. However, existing benchmarks rarely test a fundamental bottleneck whether LLMs can judge the methodological viability of a research idea before\textbf{before} investing time and computational resources. Without a reliable ``first-gate'' filter, autonomous agents risk scaling flawed methodology rather than accelerating meaningful science. We introduce SoundnessBench\textbf{SoundnessBench}, a curated benchmark of 1,099 real research proposals, with source verified extraction, reviewer-derived proxy labels, and proposal-only expert judgments on a subset of 327 proposals. Across 12 frontier LLMs, our evaluation reveals a pervasive optimism bias\textbf{optimism bias}: standard prompting yields a 74.0%74.0\% mean false-positive rate. Aggressive prompting reduces false approvals but substantially increases false rejections. Additional analyses show that optimism persists under controls for public corpus contamination, and surface features, as well as in literature-grounded agent evaluation. These results suggest a limitation in judging scientific soundness, indicating that current LLMs are not yet reliable as standalone gatekeepers for scientific rigor.

Sy-Tuyen Ho, Minghui Liu, Huy Nghiem +6

University of Maryland · All Purpose AI

AI Agent BenchmarksLLM EvaluationLLM ReliabilityAutomated Peer ReviewAI for ScienceScientific Idea Evaluation+1
AskPDF
May 28, 2026

CommunityFact: A Dynamic, Multilingual, Multi-domain Benchmark for Misinformation Detection in the Wild

Misinformation verification increasingly occurs in public, fast-moving, and multilingual online settings, where static benchmarks provide an incomplete measure of model reliability. We introduce CommunityFact, a refreshable benchmark for misinformation detection in the wild, with three major goals: coverage, granularity, and redistributability. This release contains 15,992 standalone claims across five languages and two domains. We evaluate ten LLMs under varying inference-time capabilities, including thinking and web-search. Our results show that closed-input verification remains challenging, web access yields the largest gains, and web-enabled LLMs' source-selection policies are systematically misaligned with the sources human Community Notes raters converge on -- a gap that closes through model-specific mechanisms of retrieval expansion or pruning. We further find substantial variation across language-domain slices and across the evidence ecosystems used by web-enabled systems. Beyond evaluation, CommunityFact positions Community Notes as a training signal for claim-conditioned source suggesters that could improve factual verification on novel claims.

Sahajpreet Singh, Insyirah Mujtahid, Min-Yen Kan +1

National University of Singapore

Multilingual Language Model EvaluationLLM EvaluationFake News DetectionAutomated Fact-CheckingLLM Reliability
AskPDF
May 28, 2026

Harnessing non-adversarial robustness in large language models

The work presents an approach for addressing the challenge of robustness in Large Language Models (LLMs) to alterations and potential errors caused by semantically similar but textually different prompts. Recent works have shown that these kinds of prompt variations can significantly impact the performance of LLMs on tasks. The central question is: can LLMs' robustness to semantically-neutral prompt alterations be acquired without expensive retraining of the entire model? We address this question both theoretically and through experiments. Our theoretical analysis reveals a crucial factor impacting model robustness - a systematic expected shift or perturbation-induced bias in neural network module outputs. Motivated by this analysis, we show that robustness can be achieved via a simple fine-tuning process: debiasing for robustness. We identify conditions when debiasing helps and when it does not, and demonstrate, through both theory and extensive experiments, that debiasing for robustness may indeed be a quick and efficient tool to enhance robustness and provide certification against random prompt perturbations.

Qinghua Zhou, Ellina Aleshina, Andrey Lovyagin +6

International Joint Laboratory of AI for Industry, QUST, Qingdao, China · King’s College London, London, UK · Applied AI Institute, Moscow, Russia +2

LLM Fine-TuningNeural Network RobustnessLLM ReliabilityCertified RobustnessLanguage Model RobustnessPrompt Robustness
AskPDF
May 27, 2026

Can LLMs Use Linguistic Uncertainty Markers to Reliably Reflect Intrinsic Confidence?

LLMs' linguistically expressed confidence should faithfully reflect their intrinsic uncertainty. While recent work shows LLMs struggle to use epistemic markers (e.g., "it is likely...") in a human-aligned fashion, it remains unclear whether models can apply their own linguistic confidence framework to associate markers with specific confidence levels in a stable and generalizable way, and how contextual features impact this ability. We conduct the first systematic study of this question, formalizing marker internal confidence (MIC) as the estimated intrinsic confidence a model associates with a specific epistemic marker in a given task domain. We present 7 metrics to evaluate the stability of MICs within and across distributions. Applying our analysis framework to diverse models and tasks, we find that LLMs remain faithfully miscalibrated even under model-centric interpretation of marker meanings, struggling to differentiate markers by internal confidence across distributions despite preserving a somewhat consistent ranking order across tasks. This supplies critical, complementary evidence to existing work toward a holistic understanding of faithful calibration in LLMs, emphasizing the need for more aligned and stable marker use to improve trustworthiness and reliability.

Gabrielle Kaili-May Liu, Arman Cohan

Yale University

Confidence Estimation in Language ModelsLanguage Model CalibrationLLM ReliabilityLLM Uncertainty Estimation
AskPDF
May 27, 2026

Entropy Distribution as a Fingerprint for Hallucinations in Generative Models

Large Language Models (LLMs) often generate factually incorrect outputs, commonly termed hallucinations, that undermine trust and limit deployment in high-stakes settings. Existing hallucination detection methods typically require multiple forward passes, or access to model internals. In this work, we provide theoretical background and empirical evidence that the distribution of token-level entropies, beyond the mean captured by perplexity or length-normalised entropy, serves as a fingerprint of hallucination, with distributional shape and tail behaviour carrying independent signal. We formalize hallucination detection as a statistical hypothesis test and propose the Calibrated Entropy Score (CES), a lightweight algorithm requiring only a single forward pass and black-box access to token logits. CES combines the mean signal with the maximum signal of the generated entropy through a calibrated reference CDF, producing scores that are directly comparable across models and tasks. We establish finite-sample calibration guarantees via a novel random-length Dvoretzky--Kiefer--Wolfowitz inequality, and also prove that CES detects hallucinations with probability converging to one exponentially fast in the generation length. Across eight QA benchmarks and ten generator models spanning open-source and API access models, CES achieves the highest detection performance among all single-pass black-box methods while providing formal error guarantees that existing heuristics lack. Remarkably, CES is statistically indistinguishable from multi-sample methods that require far greater computational cost, closing the gap between lightweight and expensive detection and making it suitable for real-time, large-scale deployment.

Mattia J. Villani, Pranav Deshpande, Akshay Seshadri +2

Global Technology Applied Research, JPMorganChase, New York, NY 10001, USA

Language Model Error DetectionLanguage Model Generation EvaluationHallucination DetectionHallucination in Language ModelsLLM ReliabilityLLM Hallucination Detection+1
AskPDF
May 27, 2026

The Fragility of Chain-of-Thought Monitoring Across Typologically Diverse Languages

Chain-of-thought (CoT) monitoring has been proposed as a promising safety mechanism for detecting misaligned behavior in large language models. However, its reliability remains largely unexplored beyond English and across diverse model families. We present the first large-scale evaluation of CoT monitorability across 13 diverse languages and seven frontier model families, comprising 16 models. Using adversarial-hint evaluations that require explicit intermediate computation, together with analysis of internal answer-token probabilities, we consistently find CoT unfaithfulness across languages and hint types, with an average rate of 95.9% across 8B--120B parameter models. We find that frontier models systematically exhibit strategic manipulation, including answer-switching, post-hoc rationalization, and procedural exploitation of hints, making their reasoning difficult to reliably monitor. These deceptive patterns remain especially pronounced in low-resource languages, revealing fundamental limitations in current CoT-based oversight. Our results show that CoT monitoring is fragile under linguistic distribution shift, providing a substantially weaker safety signal than English-only studies suggest. These findings motivate the development of more robust CoT monitors and complementary white-box monitoring techniques, particularly for mid- and low-resource languages. Our code is available blue{here}.

Eric Onyame, Runtao Zhou, Kowshik Thopalli +2

University of Virginia · Lawrence Livermore National Laboratory

CoT FaithfulnessMultilingual Language Model EvaluationLanguage Model Safety EvaluationDeception in Language ModelsLLM ReliabilityLanguage Model Robustness+1
AskPDF
May 26, 2026

A Paired Testing Protocol for Batch-Conditioned Refusal Robustness in LLM Serving

Safety evaluations of language models often treat serving configuration as fixed background infrastructure, but batch condition is an untested treatment variable whenever the same prompt may be evaluated alone, in a synchronized batch, or inside a continuous-batching scheduler. We synthesize four artifact-backed studies into a paired testing protocol: Study A combines local discovery, scorer-corrected adjudication, and true-batching confirmation; Study B tests cross-model generalization; Study C tests continuous-batch composition; and Study D runs a batch-invariant-kernel ablation. The local test finds safety-label changes more often than capability-label changes (0.51% vs. 0.14%), but adjudication of 63 candidate rows leaves only 17 genuine behavioral flips, implying a corrected full-set rate of 0.16%. The 15-model extension finds no detectable universal safety-over-capability skew: flips are near parity (0.94x), alignment type has no detectable association (p=0.942p=0.942, η2=0.033η^2=0.033), and output instability is the strongest tested fragility screen (r=0.909r=0.909, bootstrap 95% CI [0.65, 0.97]). In the targeted kernel ablation, standard vLLM reproduces 22/55 label flips on current score-flip candidates, while enabling VLLM_BATCH_INVARIANT=1 reduces the same test to 0/55 flips; the composition test separately finds no aggregate effect at 4.7pp sensitivity. The testing recommendation is exact-stack validation: evaluate refusal at the served batch setting, pair safety prompts with capability controls, and report low-rate directional flips separately from aggregate null effects.

Sahil Kadadekar

Independent Researcher.

LLM InferenceLanguage Model Safety EvaluationLLM Refusal BehaviorLLM ReliabilityLanguage Model RobustnessRefusal Behavior in Language Models
AskPDF
May 26, 2026

Same Answer, Different Confidence: Protocol Sensitivity in LLM Confidence Calibration

Large language models (LLM) can state their confidence in an answer (verbalized confidence), or the confidence can be read from the probability of the answer's tokens (token confidence). Comparisons of the two ask which is better calibrated, but the token probability of an answer depends on the text that precedes it. Across five open LLMs and four question-answering datasets, we fix each answer, its verbalized confidence, and its correctness label, and vary how each score is measured. The scores themselves change with the protocol, and so does the comparison between them. Changing the context in which the token probability is read shifts the gap between the two scores by about the size of the gap itself (.287 vs. .304 in expected calibration error, ECE; .189 vs. .145 in AUROC). The 95% interval of the shift excludes zero in 19 and 17 of the 20 settings, and under AUROC the shift has the same sign in all of them, so the context changes how answers are ranked. The better score flips in 8 settings under each metric, counting only flips that the intervals support, and the shift also appears in four larger models of up to 72B parameters. Most of the shift appears with changes to the instruction, not to the position of the answer. Other measurement choices also matter, and stated confidence depends on how a supplied answer is written. We turn these results into a reporting checklist for LLM confidence and its comparisons.

Hankyeol Kim, Pilsung Kang

Seoul National University

LLM EvaluationConfidence Estimation in Language ModelsLanguage Model CalibrationLLM ReliabilityLLM Uncertainty EstimationVerbalized Confidence Estimation
AskPDF
May 26, 2026

Chain-based Adaptive Reconfiguration Over Lattices for Hallucination Reduction

We introduce CAROL (Chain-based Adaptive Reconfiguration Over Lattices), a probabilistic framework for test-time hallucination reduction in large language models. Rather than relying on token-level uncertainty, CAROL defines a semantic uncertainty measure based on the consistency between generated responses and a trusted context, inducing a string-submodular objective over a lattice of textual sequences. This formulation enables hallucination mitigation to be cast as a Markov chain accept-reject process with provable convergence and near-optimality guarantees, allowing the model to iteratively refine outputs toward semantic consistency. By operating at the level of meaning, CAROL unifies hallucination detection and mitigation within a single framework. Empirical results on question answering and multi-agent reasoning benchmarks show that CAROL significantly reduces hallucinations and improves reliability and interpretability compared to likelihood-based and retrieval-augmented baselines, while maintaining competitive computational efficiency.

Joan Vendrell Gallart, Solmaz Kia, Russell Bent +1

Department of Mechanical and Aerospace University of California Irvine Irvine, CA 92617-4322, USA · T-5 Los Alamos National Laboratory Los Alamos, NM 88220, USA · CAI-4 Los Alamos National Laboratory Los Alamos, NM 88220, USA

LLM Hallucination MitigationHallucination in Language ModelsLLM ReliabilityLLM Hallucination Detection
AskPDF
May 26, 2026

How Do Uncertainty Estimators Align with Hallucination Occurrence in LLMs?

Large language models (LLMs) can hallucinate, but do model uncertainty estimators reliably reveal when they do? Although uncertainty estimators are increasingly used to identify unreliable generations, their relationship with hallucination remains poorly understood and is often taken for granted. We systematically evaluate 46 uncertainty estimators across complementary hallucination settings, covering both intrinsic hallucinations, where outputs are unsupported by the provided context, and extrinsic hallucinations, where claims are unsupported by expected model knowledge. Across four benchmarks and multiple LLMs, we find that the association between uncertainty and hallucination varies widely and is often weak, depending on the hallucination setting and the model under evaluation. These results challenge the use of uncertainty as a direct signal of hallucination and clarify when it provides useful information.

Yedidia Agnimo, Anna Korba, Annabelle Blangero +2

Ekimetrics France · Centre Inria de l’Université Grenoble Alpes France · CREST, ENSAE Institut Polytechnique de Paris France

LLM EvaluationHallucination in Language ModelsLLM ReliabilityLLM Hallucination DetectionLLM Uncertainty Estimation
AskPDF
May 26, 2026

Innovation: An Almost Characterization of Hallucination

Hallucination is a central limitation of large language models (LLMs), and substantial effort has been devoted to understanding and mitigating it. Towards this, Kalai and Vempala (STOC 2024) introduced a probabilistic framework formalizing calibration and hallucination, and showed that, with high probability, calibrated LLMs hallucinate roughly at the rate of the "missing mass", a measure of how incomplete the training data is relative to its source. This raises two fundamental questions: (i) what property of a calibrated LLM makes hallucinations unavoidable? and (ii) can hallucinations be avoided by giving up calibration? We answer these questions by introducing a simpler property we call innovation that measures the tendency of a model to produce outputs outside the training data. We show that innovation is implied by the condition for hallucination identified by Kalai and Vempala, and, further, that it is an almost characterization of hallucination: hallucination implies innovation, and conversely, innovation implies hallucination with high probability. We also provide lower bounds on the hallucination rate based on the "innovation rate", and by relating innovation rate back to missing mass, we obtain new hallucination rate lower bounds based on missing mass that extend the results of Kalai and Vempala.

Nishant P. Das, Piyush Srivastava

School of Technology and Computer Science, Tata Institute of Fundamental Research, Mumbai, Maharashtra - 400 005, India.

Hallucination in Language ModelsLanguage Model CalibrationLLM ReliabilityLanguage Modeling
AskPDF
May 26, 2026

SeDT: Sentence-Transformer Decision-Transformer Conditioning for Multi-Turn Conversation Reliability

Large language models (LLMs) achieve impressive performance when a task is fully specified in a single turn, yet the same models lose up to 39% of that performance when the identical task is revealed incrementally across multiple turns, a phenomenon documented at scale as Lost in Conversation. Crucially, this collapse is almost entirely a reliability failure; the best case, the aptitude only falls 16%, while the unreliability more than doubles (+112%). We argue that the root cause is structural, a flat conversation history assigns equal implicit weight to every prior turn, giving the model no signal to distinguish a critical constraint from incidental dialog. We present SeDT Sentence-transformer Decision-Transformer, a training-free inference-time method that resolves this by importing return-to-go conditioning from offline reinforcement learning. SeDT annotates each conversation shard with a cumulative relevance score derived from three complementary semantic, lexical, and positional signals and presents the full annotated history to the model at the final turn, without weight changes, without training data, and without discarding context. Evaluated on the Lost-in-Conversation benchmark in three LLMs and three generation tasks, SeDT outperforms the sharded baseline in all nine model-task combinations, with gains up to +37.7% in mean performance P and simultaneous reductions in unreliability in seven of the nine combinations. In short, telling the model which past turns matter is sufficient to substantially recover the performance lost in conversation.

Ramakrishna Vamsi Setti, Jagadeesh Rachapudi, Sachin Chaudhary +2

Independent Researcher · 2Drone Lab, IIT Mandi · 3UPES, Dehradun

LLM ReliabilityLanguage Model Robustness
AskPDF
May 26, 2026

The Attribution Blind Spot: Detecting When Language Models Rely on Memory Rather Than Retrieved Context

Retrieval-augmented generation promises to ground language model outputs in external evidence, yet the field has no reliable way to verify whether retrieved context actually governs generation -- a prerequisite for any high-stakes deployment. The standard assumption, that context-consistent output implies context-governed output, breaks when the retrieved document overlaps with the model's pretraining data: the model can produce faithful-looking text entirely from parametric memory, and both pathways yield indistinguishable output. We name this failure the attribution blind spot and introduce Computational Reality Monitoring (CRM) to address it. CRM operationalizes a principle adapted from cognitive science's reality monitoring framework: comparing internal representations with and without context reveals membership-conditioned representational divergence that output-level monitors systematically miss. CRM does not certify which source an individual generation used; it detects whether pretraining exposure leaves a measurable internal trajectory signature, establishing a necessary substrate for source attribution. Across nine model variants spanning three families, this divergence concentrates in architecture-specific layer patterns, receives converging support from block-level noise intervention, and generalizes across tasks and datasets while collapsing on domain-confounded benchmarks. The attribution blind spot is measurable and partially addressable: internal representations carry a diagnostic signal invisible at the output level, establishing a foundation for systems whose internal awareness of evidence provenance governs their external behavior.

Zhe Yu, Wenpeng Xing, Yunzhao Wei +4

Binjiang Institute of Zhejiang University · Zhejiang University · National Fintech Evaluation Center +2

Claim Attribution in Language ModelsMemorization in Language ModelsLLM InterpretabilityMemorization in Generative ModelsLLM ReliabilityRAG Evaluation
AskPDF
May 26, 2026

It's Not the Capability: Harness Sensitivity Is Non-Monotone Across LLM Agent Tiers

A prevalent assumption in LLM agent deployment holds that more structured harnesses universally improve reliability, and that higher-capability models need proportionally less structural guidance -- together implying a monotone inverse relationship between model capability tier and optimal harness complexity. We test this hypothesis through a controlled 432-run experiment crossing six models across four capability tiers with three harness conditions (light, balanced, strict) on HEAT-24, a 24-task synthetic benchmark with git-based workspace verification. Our results refute the monotone inverse relationship on two fronts. First, for the frontier chat model evaluated (Gemini 2.5 Flash), increased harness verbosity lowers VTSR by 29-38 percentage points -- a harness-complexity paradox. Second, for the frontier reasoning model evaluated (Qwen3.5-122B, extended thinking enabled), strict harness achieves the highest VTSR (91.7%) and the lowest latency, the opposite of the prediction. Within the constrained tier, a 2B model (Gemma4:e2B) matches strong-open-tier stability at 91.7% across all harnesses. Because each tier is represented by a single model in this study, these results should be interpreted as model-specific observations; harness sensitivity appears non-monotone across the models evaluated, and depends critically on model type (chat vs. reasoning). We introduce a six-label failure taxonomy showing that format_violation dominates capable-model failures while wrong_file dominates low-capability failures, and we derive practical tier-aware harness selection guidelines.

Yong-eun Cho

KailosLab · Seoul, Republic of Korea

AI Agent ReliabilityLLM Agent EvaluationAgent Harness OptimizationLLM ReliabilityLLM Agent HarnessesLLM Agent Reliability
AskPDF
May 25, 2026

LLM-as-a-Reviewer: Benchmarking Their Ability, Divergence, and Prompt Injection Resistance as Paper Reviewers

Large language models (LLMs) are increasingly used in academic peer review, yet their reliability, alignment with human judgment, and robustness to adversarial attacks remain poorly understood. We present a systematic benchmark of LLM-as-a-Reviewer on 898 papers stratified from NeurIPS and ICLR, evaluating 12 LLMs along three axes: rating calibration, divergence from human reviewers, and resistance to prompt injection embedded via an invisible font-mapping attack. We find that LLMs systematically overrate weaker submissions and diverge from humans in topical emphasis, under-flagging Clarity and over-flagging Reproducibility, while producing reviews two to three times longer with lower lexical diversity and a more standardized vocabulary. Prompt injection remains highly effective. Simple hidden instructions can promote low-scoring papers to acceptance-level ratings in a substantial fraction of cases, with effectiveness varying sharply across model families. While LLMs offer utility in structuring evaluations, their integration into peer review requires safeguards against both intrinsic biases and adversarial risks.

Lingyao Li, Junjie Xiong, Changjia Zhu +5

University of South Florida · Missouri University of Science and Technology · University of Alabama +3

LLM EvaluationLLM-as-a-JudgeAdversarial Attacks on LLMsLLM ReliabilityLanguage Model Bias EvaluationAutomated Peer Review
AskPDF
May 25, 2026

ARBITER: Reasoning Trajectory Basins and Majority Vote Failures in Test-Time Sampling

When language models use test-time sampling, they generate multiple reasoning trajectories and select an answer by majority vote. We show that these trajectories are not independent: for a given question, they concentrate into a small number of clusters, or reasoning basins, each defined by a normalized final answer and the solutions that reach it. A majority vote therefore selects the most stable basin rather than the most accurate one, which creates wrong-majority failures where the correct answer is present but outvoted. We introduce ARBITER, a model-agnostic approach that models interactions between basins using only the base model's own sampled outputs, hidden states, and derived evidence. Most direct correction strategies fail; ARBITER instead uses conservative additive evidence on top of consensus. In its simplest parameter-free form, ARBITER-Δ adds same-model evidence to the majority prior, while ARBITER-Enc augments this with bounded residual signals from hidden states over complete solutions. On GSM8K with Qwen3-4B, consensus over K=24 samples achieves around the mid-94% range, while a same-pool top-2 oracle reaches around the mid-96% range. ARBITER recovers a subset of these cases using zero external information. Across three model families and three math benchmarks, it yields consistent gains with no net-negative cases; for example, on Llama-3.1-8B MMLU-HS-Math, it improves accuracy from the mid-78% range to the mid-82% range, recovering about 22% of the available oracle headroom, indicating that this headroom can be partially recovered from the sample pool itself.

Meng Cai, Lars Kulik, Farhana Choudhury

School of Computing and Information Systems University of Melbourne

Self-Consistency DecodingTest-Time ScalingLLM ReliabilityLLM ReasoningMajority Voting
AskPDF
May 24, 2026

Trust but Verify: Prover-Verifier Deliberation for Selective LLM Prediction

Reliably knowing when a language model is correct is almost as important as being correct. We introduce prover-verifier deliberation (PVD), an inference-time protocol grounded in interactive proof theory, as a mechanism for selective prediction: the protocol produces both an answer and a structured confidence verdict, allowing a system to report high-confidence answers while abstaining on uncertain cases. In each dialogue, a prover defends a candidate answer through checkable sub-claims while a verifier issues targeted challenges and returns Accept, Challenge, or Reject. Because frozen language models are imperfect provers and verifiers operating over a noisy channel, formal soundness and completeness guarantees do not transfer; instead, we characterize the protocol empirically through its coverage-precision behavior. Our main experiment uses Claude Sonnet 4.6 as prover and Claude Haiku 4.5 as verifier on GPQA Diamond. Questions accepted with no answer revision, which we call Accept + No Change (ANC), are reported as the high-confidence subset; we evaluate this subset by its precision and coverage. ANC separates reliable from unreliable answers, yielding a ∼\sim30pp HC-Prec gap over the non-ANC complement. Robustness experiments with GPT and Gemini pairings show that high HC-Prec can transfer across model families, while verifier strictness and domain competence largely determine the size of the selection gap. On Humanity's Last Exam, weaker prover-verifier pairings can collapse or invert the ANC signal, illustrating a practical failure mode when the verifier operates outside its effective region. Comparisons with self-consistency, universal self-consistency, multi-agent debate, and Reflexion suggest that prover-verifier deliberation supplies a distinct argument-defensibility signal for selective prediction.

João Sedoc, Baotong Zhang, Dean Foster

New York University

LLM Answer VerificationSelective PredictionConfidence Estimation in Language ModelsSelective InferenceLLM ReliabilityLLM Uncertainty Estimation+1
AskPDF
May 24, 2026

Faithfulness Metrics Don't Measure Faithfulness: A Meta-Evaluation with Ground Truth

Chains of thought (CoTs) have become central in interpreting and auditing behaviors of large language models. Yet growing evidence suggests that these traces often fail to faithfully represent the computations behind a model's predictions. Several faithfulness metrics have been proposed, but whether they indeed measure faithfulness remains unknown. Answering this requires ground-truth labels, which are hard to obtain since internal computations are not directly observable. Consequently, most works proposing metrics report only absolute scores or comparisons to prior metrics, and the few existing benchmarks rely on proxies like plausibility or importance, properties orthogonal to faithfulness that can mislead about whether a CoT can be trusted. We address this challenge by constructing tasks whose outputs reveal which intermediate computations must have produced them, and developing an automated labeling pipeline that yields ground-truth faithfulness labels at both the step and CoT level. Building on this methodology, we present BonaFide, a benchmark of 3,066 labeled CoTs across 13 tasks and 10 models, and use it to conduct the first systematic evaluation of prominent faithfulness metrics. Our experiments show that most metrics perform near chance, exhibit strong prediction biases and degrade on longer CoTs. The best metric reaches only 0.70 AUROC at the CoT level while another reaches 0.59 at the step level, with neither transferring across settings, while entailing prohibitively high computational cost. Our results expose fundamental gaps in current faithfulness evaluation and call for the development of more reliable and efficient metrics.

Yoav Gur-Arieh, Ana Marasović, Mor Geva

Tel Aviv University · University of Utah

CoT FaithfulnessLLM EvaluationLanguage Model Generation EvaluationFaithfulness of Language Model ExplanationsLLM Reliability
AskPDF
May 21, 2026

Graph Alignment Topology as an Inductive Bias for Grounding Detection

Large Language Models (LLMs) are optimized to produce distributionally plausible continuations rather than to explicitly verify whether generated propositions are entailed by source documents. This inductive bias enables generalization, but it does not encode whether responses are grounded with respect to a reference. These issues limit the use of LLMs in domains where strict factual correctness is crucial, such as clinical decision support. Existing hallucination detection approaches improve factuality through retrieval augmentation, self-consistency, or claim verification, but generally do not learn directly over alignment topology. To leverage alignment topology as an inductive bias, we construct aligned bipartite graphs between reference information and LLM outputs and train a graph neural network (GNN) to model alignment structure using message passing. The method achieves state-of-the-art results on four diverse hallucination and question-answering datasets, outperforming all compared methods, including foundational LLMs such as GPT-4o.

Paul Landes, Pranav Herur, Adam Cross +1

†Department of Pediatrics, University of Illinois College of Medicine Peoria · ♣University of Illinois Urbana-Champaign · ♠Siebel School of Computing and Data Science, University of Illinois Urbana-Champaign +1

LLM GroundingHallucination in Language ModelsLLM ReliabilityLLM Hallucination Detection
AskPDF
May 21, 2026

Hallucination as Commitment Failure: Larger LLMs Misfire Despite Knowing the Answer

Hallucination is often viewed as a direct consequence of missing knowledge: a model answers incorrectly when the correct answer is absent from its generation-time distribution, and correctly when it is present. We test this assumption by introducing a semantic notion of answer availability that aggregates token-level variants expressing the same answer concept, and asks whether the correct concept is already available at the moment the model commits to an answer. Across Qwen and Llama models from 0.8B to 72B in both Instruct and Base variants, 16-47% of Instruct hallucinations occur with substantial probability mass already on the correct concept, and the rate rises monotonically with scale. Comparing such failures against correct generations with matched semantic support, the distinguishing factor is not whether the correct concept is represented, but how its probability is distributed: correct generations concentrate mass on a single surface form, hallucinations disperse it across alternatives. The same sharpening asymmetry extends across multi-token generation and is detectable in pre-generation hidden states. Together, these results identify a single mechanism: instruction tuning sharpens answer commitment with scale, making helpfulness and confident hallucination two consequences of the same underlying disposition.

Jewon Yeom, Jaewon Sok, Heejun Kim +3

Graduate School of Data Science, Seoul National University · Department of Rural Systems Engineering, Seoul National University · Electrical Engineering and Computer Science, Gwangju Institute of Science and Technology +1

LLM InterpretabilityHallucination in Language ModelsLLM ReliabilityInstruction Tuning
AskPDF
May 20, 2026

Calibration vs Decision Making: Revisiting the Reliability Paradox in Unlearned Language Models

Machine unlearning aims to remove the influence of specific training data from a model while preserving reliable behavior on the remaining data, making reliable prediction and uncertainty estimation essential for evaluation. Calibration is commonly used as a proxy for reliability in language models, but low calibration error does not necessarily imply reliable decision rules, as models may rely on spurious correlations while remaining well calibrated. We investigate this gap in generative language models using the multiple-choice question-answering evaluation protocol on the TOFU benchmark, measuring probabilistic reliability with calibration metrics (ECE, MCE, Brier) and decision-rule reliability via attribution-based shortcut detection with Integrated Gradients and Local Mutual Information. We find that fine-tuned models achieve low calibration error (ECE ~ 0.04) compared to pretrained models (ECE > 0.5), and models after unlearning retain similarly low calibration despite reduced accuracy on the forget split, while attribution analysis shows increased reliance on correlation-based tokens. These results demonstrate that good calibration can coexist with shortcut-based decision rules after unlearning, extending the reliability paradox to the machine unlearning setting.

Divyaksh Shukla, Ashutosh Modi

Indian Institute of Technology Kanpur (IIT Kanpur)

Shortcut Learning in Language ModelsConfidence Estimation in Language ModelsLanguage Model CalibrationLLM ReliabilityLLM UnlearningMachine Unlearning Evaluation
AskPDF
May 20, 2026

Do No Harm? Hallucination and Actor-Level Abuse in Web-Deployed Medical Large Language Models

Medical large language models (LLMs), including custom medical GPTs (MedGPTs) and open-source models, are increasingly deployed on web platforms to provide clinical guidance. However, they pose risks of hallucination, policy noncompliance, and unsafe design. We conduct a large-scale assessment of 6,233 MedGPTs, evaluating a stratified sample of 1,500, together with 10 open-source LLMs. We introduce two frameworks: MedGPT-HEval for hallucination detection and an LLM-based pipeline for assessing policy violations and developer intent. Our results show that 25-30% of MedGPTs exhibit low factual accuracy, with bottom- and middle-tier models at highest risk; 33.6-54.3% violate operational thresholds, and 57.06% of Action-enabled models lack adequate privacy disclosures. Compared with open-source models, MedGPTs achieve higher factual accuracy and semantic alignment, though open-source models are more stable. These results reveal systemic gaps in hallucination and compliance, highlighting the need for multi-metric evaluation and stronger safeguards. We release HAA-MedGPT, a structured dataset that supports future research on the safety of web-facing medical LLMs.

Sunday Oyinlola Ogundoyin, Muhammad Ikram, Rahat Masood

Macquarie University, Sydney, Australia · The University of New South Wales, Sydney, Australia

Language Model Safety EvaluationHealthcareLLM AuditingHallucination in Language ModelsLLM ReliabilityLLM Hallucination Detection+2
AskPDF
May 19, 2026

Reading Calibrated Uncertainty from Language Model Trajectories

The maximum softmax probability (MSP) represents a default approach when evaluating uncertainty quantification for language model generation with structured output. Although cheap, it is often miscalibrated. Methods that probe the model's internal activations feed raw hidden states into opaque classifiers, reading activations as static snapshots and leaving implicit the layer-wise trajectory by which a representation is formed. Yet, similar endpoints can arise from very different paths, and how evidence accumulates, reinforces, or reverses across depth might reveal uncertainty that final probabilities obscure. We extract eleven scale-invariant geometric features, tracing the cumulative path of per-layer MLP updates, and feed them to a sparse linear probe. The probe outperforms MSP under selective abstention, with gains scaling with baseline miscalibration up to 21 AURC points. Because every feature has a closed-form geometric meaning, the probe's coefficients trace how and where along depth errors take shape -- which layers commit prematurely, which contradict the running state, where trajectories drift away from their endpoint.

Aliai Eusebi, Alexander Herzog, Xiaoyu Liang +3

University College London, London, United Kingdom · Core46, London, United Kingdom

Confidence Estimation in Language ModelsLLM ReliabilityLLM Uncertainty EstimationRepresentation Geometry in Language ModelsUncertainty CalibrationLanguage Model Probing
AskPDF
May 19, 2026

BalanceRAG: Joint Risk Calibration for Cascaded Retrieval-Augmented Generation

Large language models (LLMs) can enhance factuality via retrieval-augmented generation (RAG), but applying RAG to every query is unnecessary when the model-only answer is reliable. This motivates cascaded RAG: each query is first handled by an LLM-only branch, escalated to a RAG fallback only if the primary branch is uncertain, and abstained from when neither branch is sufficiently trustworthy. However, calibrating such cascades stage by stage may be conservative, since the final utility depends on joint uncertainty thresholding of LLM-only and RAG. In this work, we develop BalanceRAG to certify threshold pairs at a target risk level. Given uncertainty scores from the two branches, BalanceRAG frames each threshold pair as an operating point on a two-dimensional lattice and identifies safe operating points using sequential graphical testing. This enables risk-adaptive threshold calibration, controlling the system-level error rate among accepted points, while retaining more examples. Furthermore, BalanceRAG extends to multi-risk calibration, allowing retrieval usage to be bounded together with the selection-conditioned risk. Experiments on three open-domain question answering (QA) benchmarks across multiple LLM backbones demonstrate that BalanceRAG meets prescribed risk levels, preserves higher coverage and more accepted correct examples, and reduces unnecessary retrieval calls compared with always-on RAG.

Zijun Jia, Yuanchang Ye, Sen Jia +6

Beihang University · Zhejiang University of Finance & Economics · Shenzhen Institute of Advanced Technology +1

Retrieval-Augmented GenerationSelective PredictionLanguage Model CalibrationLLM Reliability
AskPDF