{"slug": "dead-text-or-binding-clause-measuring-and-restoring-constraint-influence-in-box", "title": "Dead text or binding clause? Measuring and restoring constraint influence in black-box LLM dialogues", "summary": "A new arXiv preprint (2608.12599v1) introduces a method to measure and restore constraint influence in black-box LLM dialogues, finding that behavioral relapse—models continuing to enact revoked constraints—increases with constraint load at an 8B operating point, while stronger models remain at floor. The method, using a contract ledger, sequential ablation probe, and repair ladder, significantly reduces relapse against a no-ledger baseline (95% CI, p < 0.05), with a one-sentence tombstone note recovering about a third of the effect. The probe predicts relapse before delivery (AUROC 0.85), and the approach operates at 1.2x delivery overhead and 0.3% of API compute.", "body_md": "arXiv:2608.12599v1 Announce Type: new\nAbstract: Multi-turn dialogues let users revoke constraints as easily as impose them, but revocation does not reliably take effect: models keep enacting withdrawn requirements (occasionally beneath comments asserting their removal), a failure we call \\emph{behavioral relapse}, or revocation inertia. No existing instrument measures this influence per clause, predicts it before delivery, or repairs it under matched budgets. \\sysname{} closes the three gaps through the model API alone: a contract ledger pairs every constraint with an executable checker, records revocations as tombstones, and compiles the net constraint state ahead of time into a single specification; a sequential ablation probe measures per-clause adherence and incremental behavioral effect; a repair ladder operates under token- and attempt-matched budgets. On \\dataname{} (\\NTasks{} HumanEval tasks, \\NClauses{} verified checkers), relapse at an 8B operating point climbs from \\ScaleDelayedMTwo{} to \\ScaleDelayedMEight{} as constraint load grows, while stronger models sit at floor. Under matched checkers, model, and budget, ahead-of-time compilation significantly reduces relapse against a no-ledger verifier-retry baseline (\\RestoreDiff{}, 95\\% CI \\RestoreDiffCI{}, $p$ \\RestoreDiffP{}); adaptive ladder interventions stacked on top add no detectable gain (95\\% confidence excludes gains $\\geq$ \\LadderExcludedGain{}). The probe predicts relapse before delivery (AUROC \\AurocPrimary{}); a one-sentence tombstone note recovers about a third of the compilation effect and survives a placebo control. At \\CostDeliveryFactor{} delivery overhead and \\CostTotalHedged{} of API compute for every result, revocation failure becomes a measurable, predictable, and repairable property of dialogue state rather than an invisible one.", "url": "https://wpnews.pro/news/dead-text-or-binding-clause-measuring-and-restoring-constraint-influence-in-box", "canonical_source": "https://arxiv.org/abs/2608.12599", "published_at": "2026-08-14 04:00:00+00:00", "updated_at": "2026-08-14 04:10:34.201954+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-safety", "ai-research"], "entities": ["arXiv", "HumanEval"], "alternates": {"html": "https://wpnews.pro/news/dead-text-or-binding-clause-measuring-and-restoring-constraint-influence-in-box", "markdown": "https://wpnews.pro/news/dead-text-or-binding-clause-measuring-and-restoring-constraint-influence-in-box.md", "text": "https://wpnews.pro/news/dead-text-or-binding-clause-measuring-and-restoring-constraint-influence-in-box.txt", "jsonld": "https://wpnews.pro/news/dead-text-or-binding-clause-measuring-and-restoring-constraint-influence-in-box.jsonld"}}