How Language Models Fail: Token-Level Signatures of Committed and Persistent Reasoning Failures

• URL: http://arxiv.org/abs/2606.06635v1
• Date: Thu, 04 Jun 2026 18:36:42 GMT
• Title: How Language Models Fail: Token-Level Signatures of Committed and Persistent Reasoning Failures
• Abstract summary: Failures in language model reasoning emerge through distinct processes that leave identifiable signatures in the reasoning trace.<n>The first is committed failure, in which a model locks onto an incorrect reasoning path early in its trace.<n>In the second, persistent uncertainty instead accumulates throughout, and the full trace is needed to best distinguish failing from successful completions.
• Score: 21.666726162280696
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Failures in language model reasoning emerge through distinct processes that leave identifiable signatures in the reasoning trace. We characterize these failures using token-level uncertainty signals, finding they arise through two empirically distinguishable processes. The first is committed failure, in which a model locks onto an incorrect reasoning path early in its trace. A central diagnostic signature is the commitment point, beyond which considering additional tokens hurt rather than help failure detection. In the second, persistent uncertainty, uncertainty instead accumulates throughout, and the full trace is needed to best distinguish failing from successful completions. These signatures reproduce across 23 model-dataset configurations, with the framework's falsifiable predictions holding in 20 of 23 cases, well above chance across both failure modes. Finally, we demonstrate our failure mode framework has direct implications for self-consistency, identifying when uncertainty signals complement it and when it can be selectively skipped. These results offer a foundation for understanding when LLM reasoning failures become detectable and for adapting detection strategies accordingly.

Related papers

• DCFA: Dual-view Causal-inspired Attribution for Failure Reasoning in LLM-based Multi-agent Systems [61.11435469308521]
  We propose DCFA, a training-free framework for failure attribution in large language model systems.<n> DCFA integrates a global module that constructs structured causal-inspired dependency graphs from system traces.<n>Experiments show that DCFA improves step-level accuracy by up to 8.27% over state-of-the-art baselines.
  arXiv  Detail & Related papers  (2026-09-04T05:29:31Z)
• Wrong Prediction, Right Answer: Recovering Evidence from Collapsed LLM Sequence Scores [31.709549159768727]
  We observe a consistent readout gap across diverse reasoning benchmarks.<n> hidden-state probes successfully decode correct answers even when native sequence scoring completely collapses.<n>Our results show that many apparent zero-shot reasoning deficits are expression failures masking intact internal logic.
  arXiv  Detail & Related papers  (2026-08-31T16:43:55Z)
• From Uncertainty to Failure Attribution: Self-Diagnosing Models for Failure Attribution under Distribution Shift [0.3948325938742681]
  We introduce a problem setting for failure attribution under distribution shift.<n>This enables the models not only to detect OOD samples, but also to find out the reason for their failure.<n>The solution we propose is called self-diagnosing models, which are capable of jointly learning predictive output, predictive uncertainty, and a failure attribution signal.
  arXiv  Detail & Related papers  (2026-08-08T06:32:20Z)
• Escaping Confidence Trap: Evolutionary Decoding for Mathematical Reasoning in Diffusion LLMs [69.52853771994451]
  We analyze the decoding trajectories of LLaDA 2.0 and identify a recurring diffusion confidence trap.<n>We propose Evolutionary Decoding, a training-free test-time scaling framework that views diffusion decoding as an evolutionary process.
  arXiv  Detail & Related papers  (2026-08-01T11:48:25Z)
• The Knowing-Saying Gap: When Probes See Errors that Confidence Misses [20.46211358172975]
  Linear probes detect corrupted context in language models with near-perfect accuracy, yet this does not translate into reliable failure prediction.<n>Across multi-hop arithmetic chains, probes that detect corruption turn out to be uninformative about final answer correctness.<n>As a real-time monitor, probe-based interventions are sharply model and error-type dependent.
  arXiv  Detail & Related papers  (2026-07-21T12:13:10Z)
• Does It Fail to See or Fail to Know? Attributing Errors in Vision-Language Models [33.564453692157095]
  Uncertainty should not only indicate whether the model is likely to fail but also diagnose why it is uncertain.<n>We propose a unified framework for disentangling failure modes and investigate whether pre-generation signals can predict these failure sources.<n>Our main finding is that recognition-related failures are best captured by visual-token representations, while failures that remain after recognition are better captured by prompt-conditioned hidden states.
  arXiv  Detail & Related papers  (2026-07-06T05:11:45Z)
• How LLMs Detect and Correct Their Own Errors: The Role of Internal Confidence Signals [6.467495925520036]
  Large language models can detect their own errors and sometimes correct them without external feedback.<n>We investigate this through the lens of second-order models of confidence from decision neuroscience.
  arXiv  Detail & Related papers  (2026-04-24T06:33:32Z)
• Silent Commitment Failure in Instruction-Tuned Language Models: Evidence of Governability Divergence Across Architectures [0.0]
  We introduce governability -- the degree to which a model's errors are detectable before output commitment and correctable once detected.<n>We show benchmark accuracy does not predict governability, correction capacity varies independently of detection, and identical governance scaffolds produce opposite effects across models.<n>We propose a Detection and Correction Matrix classifying model-task combinations into four regimes: Governable, Monitor Only, Steer Blind, and Ungovernable.
  arXiv  Detail & Related papers  (2026-03-22T21:50:28Z)
• Certainty-Validity: A Diagnostic Framework for Discrete Commitment Systems [0.0]
  "Machine learning evaluation metrics" assume that all errors are equivalent discrete commitment systems.<n>"ConfidentIncorrect" behavior is where models hallucinate structure in ambiguous data.<n>"Good training" for reasoning systems must be defined not by accuracy but by maximizing the Certainty-Validity Score.
  arXiv  Detail & Related papers  (2026-02-10T21:53:02Z)
• Decomposed Prompting Does Not Fix Knowledge Gaps, But Helps Models Say "I Don't Know" [47.930782177987446]
  Large language models often struggle to recognize their knowledge limits in closed-book question answering, leading to confident hallucinations.<n>We evaluate three task-equivalent prompting regimes: Direct, Assistive, and Incremental, across different model scales and multi-hop QA benchmarks.<n>Because factual knowledge is stable while hallucinations are agreement, cross-regime provides a precise signal of internal uncertainty.
  arXiv  Detail & Related papers  (2026-02-04T18:39:58Z)
• Hierarchical Vision Language Action Model Using Success and Failure Demonstrations [60.82332413442677]
  We introduce VINE, a hierarchical vision-language-action model that separates high-level reasoning from low-level control.<n>System 2 performs feasibility-guided tree search over a 2D scene-graph abstraction.<n>System 1 executes low-level actions without modifying the agent's core skills.
  arXiv  Detail & Related papers  (2025-12-03T15:58:38Z)
• CLUE: Non-parametric Verification from Experience via Hidden-State Clustering [64.50919789875233]
  We show that correctness of a solution is encoded as a geometrically separable signature within the trajectory of hidden activations.<n>ClUE consistently outperforms LLM-as-a-judge baselines and matches or exceeds modern confidence-based methods in reranking candidates.
  arXiv  Detail & Related papers  (2025-10-02T02:14:33Z)
• LoGU: Long-form Generation with Uncertainty Expressions [49.76417603761989]
  We introduce the task of Long-form Generation with Uncertainty(LoGU)<n>We identify two key challenges: Uncertainty Suppression and Uncertainty Misalignment.<n>Our framework adopts a divide-and-conquer strategy, refining uncertainty based on atomic claims.<n>Experiments on three long-form instruction following datasets show that our method significantly improves accuracy, reduces hallucinations, and maintains the comprehensiveness of responses.
  arXiv  Detail & Related papers  (2024-10-18T09:15:35Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.