{"slug": "verytrace-verifying-reasoning-traces-through-compilable-formalism-and-structured", "title": "VeryTrace: Verifying Reasoning Traces through Compilable Formalism and Structured Verification", "summary": "Researchers introduced VeryTrace, a zero-shot verification-and-repair framework that formalizes Chain-of-Thought reasoning traces into a compilable Domain-Specific Language, enabling step-level error detection and correction. Across competition math, robotics planning, and kinship reasoning tasks, VeryTrace improved accuracy over baselines without domain-specific training, demonstrating that structured verification enhances LLM reasoning reliability.", "body_md": "arXiv:2606.24124v1 Announce Type: new\nAbstract: Multi-step reasoning with Chain-of-Thought (CoT) prompting remains fragile: logical errors or hallucinations in early steps silently propagate, producing confident but incorrect conclusions. This paper presents VeryTrace, a zero-shot verification-and-repair framework that formalizes natural-language reasoning traces into a structured, compilable representation. VeryTrace introduces a Domain-Specific Language (DSL) that (i) makes step dependencies explicit, (ii) mechanizes quantitative content as executable expressions, and (iii) structures semantic inferences via deduction schemas. Our hybrid verifier combines deterministic checks for computational correctness, dependency resolution, and constraint satisfaction with targeted LLM audits for non-mechanizable semantic judgments, enabling step-level error localization and repair.\nAcross three diverse domains-competition mathematics (AIME 2025), robotics planning (LLM-BabyBench), and kinship reasoning (CLUTRR), VeryTrace improves accuracy over zero-shot baselines on state-of-the-art LLMs without requiring domain-specific training or in-context examples, demonstrating that formalized trace verification achieves both precision and generalization.", "url": "https://wpnews.pro/news/verytrace-verifying-reasoning-traces-through-compilable-formalism-and-structured", "canonical_source": "https://arxiv.org/abs/2606.24124", "published_at": "2026-06-24 04:00:00+00:00", "updated_at": "2026-06-24 04:30:41.138547+00:00", "lang": "en", "topics": ["large-language-models", "ai-research", "ai-safety", "natural-language-processing"], "entities": ["VeryTrace", "Chain-of-Thought", "AIME 2025", "LLM-BabyBench", "CLUTRR"], "alternates": {"html": "https://wpnews.pro/news/verytrace-verifying-reasoning-traces-through-compilable-formalism-and-structured", "markdown": "https://wpnews.pro/news/verytrace-verifying-reasoning-traces-through-compilable-formalism-and-structured.md", "text": "https://wpnews.pro/news/verytrace-verifying-reasoning-traces-through-compilable-formalism-and-structured.txt", "jsonld": "https://wpnews.pro/news/verytrace-verifying-reasoning-traces-through-compilable-formalism-and-structured.jsonld"}}