{"slug": "valcon-benchmarking-and-enhancing-value-consistency-in-large-language-models-via", "title": "VALCON: Benchmarking and Enhancing Value Consistency in Large Language Models via Dual-Role Variation", "summary": "Researchers at Tongji University's TJUNLP lab released VALCON, a role-driven benchmark framework that tests whether large language models keep consistent values when switching between generating actions and evaluating them, evaluated across 28 LLMs. The team reports pervasive cross-role inconsistency, finding that models with highly compliant outputs frequently fail to align their judgments with their own generations, and traces the gap to a role-dependent representational decoupling. VALCON provides three benchmark suites (VAB-Consistency, VAB-Boundary, VAB-Challenge) plus a representation-directed intervention that mitigates the consistency failures, with code and data at https://github.com/tjunlp-lab/VALCON.", "body_md": "arXiv:2610.21304v1 Announce Type: new \nAbstract: Current value alignment evaluations for large language models (LLMs) heavily rely on output compliance, overlooking whether LLMs maintain value consistency when alternating between generator and evaluator roles. Such consistency is increasingly important because modern LLM agents and alignment pipelines often rely on the same LLM to both generate actions and evaluate their acceptability. To mitigate this limitation, we propose VALCON, a role-driven data construction framework that operationalizes value consistency into three benchmark suites (VAB-Consistency, VAB-Boundary, VAB-Challenge) via dual-role variation. By systematically pairing an LLM's generative behavior with its evaluative judgments under identical principles, VALCON synthesizes target evaluation data that exposes hidden value discrepancies. Experiments on 28 LLMs reveal pervasive cross-role inconsistency: models with highly compliant outputs frequently fail to align their judgments with their own generations. To understand and address this gap, we perform mechanistic probing that uncovers a role-dependent representational decoupling associated with this behavioral asymmetry. Guided by these insights, we propose a representation-directed intervention that effectively mitigates consistency failures. Our work establishes value consistency as a distinct alignment dimension and provides a complete evaluation-analysis-mitigation pipeline. Code and data are available at https://github.com/tjunlp-lab/VALCON.", "url": "https://wpnews.pro/news/valcon-benchmarking-and-enhancing-value-consistency-in-large-language-models-via", "canonical_source": "https://www.machinebrief.com/news/valcon-benchmarking-and-enhancing-value-consistency-in-large-x2dd", "published_at": "2026-10-16 04:00:00+00:00", "updated_at": "2026-10-16 05:09:10.580472+00:00", "lang": "en", "topics": ["large-language-models", "ai-safety", "ai-research", "ai-agents"], "entities": ["VALCON", "TJUNLP", "Tongji University", "VAB-Consistency", "VAB-Boundary", "VAB-Challenge"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/valcon-benchmarking-and-enhancing-value-consistency-in-large-language-models-via", "markdown": "https://wpnews.pro/news/valcon-benchmarking-and-enhancing-value-consistency-in-large-language-models-via.md", "text": "https://wpnews.pro/news/valcon-benchmarking-and-enhancing-value-consistency-in-large-language-models-via.txt", "jsonld": "https://wpnews.pro/news/valcon-benchmarking-and-enhancing-value-consistency-in-large-language-models-via.jsonld"}}