mlsanityData to 7 Oct 2026

Probabilistic Soundness Guarantees in LLM Reasoning Chains

arXiv 2507.12948 ↗First posted 17 Jul 2025cs.LG cs.CLcited by 13

On the map: LLM reasoning → Test-time compute and efficient thinking

Abstract (arXiv)

In reasoning chains generated by large language models (LLMs), initial errors often propagate and undermine the reliability of the final conclusion. Current LLM-based error detection methods often fail to detect propagated errors because earlier errors can corrupt judgments of downstream reasoning. To better detect such errors, we introduce Autoregressive Reasoning Entailment Stability (ARES), a probabilistic framework that evaluates each reasoning step based solely on previously-verified premises. This inductive method yields a nuanced score for each step and provides certified statistical guarantees of its soundness, rather than a brittle binary label. ARES achieves state-of-the-art performance across four benchmarks (72.1% Macro-F1, +8.2 points) and demonstrates superior robustness on very long synthetic reasoning chains, where it excels at detecting propagated errors (90.3% F1, +27.6 points).

Key facts. Cited by 13 papers in the map (data through 7 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 3 of its 55 references that are in the map. First posted on arXiv 17 Jul 2025.

Builds on

3 of 55 references.

References · 3

  1. 2025Qwen3 Technical Reportcited by 8,195
  2. 2024Faithfulness vs. Plausibility: On the (Un)Reliability of Explanations from Large Language Modelscited by 104
  3. 2025Evaluating Step-by-step Reasoning Traces: A Surveycited by 62

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 5; 2026: 8.

13 papers.

Sort:

By area:

  1. 2026Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Modelscited by 22
  2. 2025E-valuator: Reliable Agent Verifiers with Sequential Hypothesis Testingcited by 14
  3. 2025SSR: Socratic Self-Refine for Large Language Model Reasoningcited by 6
  4. 2026Interpretability Can Be Actionablecited by 5
  5. 2026Limited Reasoning Space: The cage of long-horizon reasoning in LLMscited by 3
  6. 2025Prototyping Multimodal GenAI Real-Time Agents with Counterfactual Replays and Hybrid Wizard-of-Ozcited by 2
  7. 2025T-FIX: Text-Based Explanations with Features Interpretable to eXpertscited by 2
  8. 2026Probabilistic Programs of Thoughtcited by 1
  9. 2025When LRP Diverges from Leave-One-Out in Transformerscited by 0
  10. 2026Missingness Bias Calibration in Feature Attribution Explanationscited by 0
  11. 2026DanceHA: A Multi-Agent Framework for Document-Level Aspect-Based Sentiment Analysiscited by 0
  12. 2026Shattering the Autoregressive Curse: Dynamic Epistemic Entropy Orchestrated Erasable Reinforcement Learning for LLMscited by 0
1 more
  1. 2026Online Safety Monitoring for LLMscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Sample Smart, Not Hard: Correctness-First Decoding for Better Reasoning in LLMs
  2. 2026"I May Not Have Articulated Myself Clearly": Diagnosing Dynamic Instability in LLM Reasoning at Inference Time
  3. 2026Beyond Confidence: Stability-Aware Test-Time Adaptation for LLM Reasoning
  4. 2026ChainUQ: Reasoning Consistency-Aware Uncertainty Quantification for Large Language Models
  5. 2025Grammars of Formal Uncertainty: When to Trust LLMs in Automated Reasoning Tasks
  6. 2025Investigating the Robustness of Deductive Reasoning with Large Language Models
  7. 2026CITE: Anytime-Valid Statistical Inference in LLM Self-Consistency
  8. 2025ReasonBENCH: Benchmarking the (In)Stability of LLM Reasoning