Latent Dynamics · 2026-10-14
Failure Recovery via Belief Updates and World Modeling for Long-Horizon Agents
Xiyang Wu, Zhichao Liu, Shengxin zhang, Zongxia Li, Ming Li, Tianyi Zhou, Dinesh Manocha
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
BeliefWeaver is a long-horizon failure-recovery framework that keeps a persistent relational belief state over repair frontiers across rollback, and uses an implicit world model (IWM) to evaluate repair hypotheses and prospective progress. Execution is gated by diagnostic evidence and ranked by belief. The framework is distilled into a 9B student via SFT and on-policy distillation (OPD) to reduce teacher inference reliance.
Why it matters
The authors target failures caused by decisions made several steps earlier, where local reflection/backtracking is insufficient. BeliefWeaver’s persistent evidence across attempts and prospective reasoning aim to reprioritize related repair directions beyond the latest attempt. Distilling IWM reasoning into a 9B student is presented as a way to keep recovery improvements while reducing teacher-model inference cost.
Method
- Maintain a persistent relational exploration graph over frontiers, attempts, observations, and checkpoints; outcomes update frontier priorities across rollback via unary evidence factors; diagnostic evidence gates execution eligibility.
- Use an implicit world model (IWM) to propose/evaluate candidate repair frontiers: evidence judgment (consistent/insufficient/contradicted), prospective evaluation (repair progress and frontier relations), and relation inference to connect related hypotheses.
- Distill recurring IWM roles into a 9B student using supervised fine-tuning (SFT) and on-policy distillation (OPD), keeping the recovery controller/executor and belief-update rules unchanged for student execution.
Abstract (from arXiv)
Long-horizon agents often struggle to recover from failures whose causes lie several steps earlier, beyond what local reflection or backtracking can resolve. We introduce BELIEFWEAVER, a framework that combines persistent evidence across attempts with prospective reasoning over repair directions. It maintains a relational belief state over frontiers, attempts, observations, and checkpoints. Execution evidence updates related frontier priorities across attempts and rollback, while diagnostic evidence gates execution. An implicit world model (IWM) evaluates repair hypotheses, predicts progress, and identifies relations among frontiers to guide exploration. We further distill these recurring IWM reasoning components into a 9B student through supervised fine-tuning and on-policy distillation (OPD). Across three terminal-based benchmarks, BELIEFWEAVER improves aggregate final score by 5.1 percentage points over native Terminus-2 continuation, with positive trends in exploration efficiency. The distilled 9B student improves final score by 9.8 percentage points over the reported native baseline while reducing estimated teacher API cost by 54.8% relative to the matched teacher. These results show that persistent evidence and prospective reasoning can improve recovery in long-horizon agent tasks while substantially reducing reliance on teacher-model inference.
Related papers
- EvoBelief: Falsifiable World Models for Self-Improving LLM Agents
- World Models Dream of Success: Diagnosing and Repairing Failure Insensitivity in Robot World Models
- ANTE: Anticipate, Then Execute -- Learning Internal World Models for Coding Agents from Verified Beliefs
- Memento 3: Model-Based Recursive Self-Improvement through Reflective Rulebooks
- Safe, Persistent, and Evolving Agent Harness for Understanding Partially Observable Worlds
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-13 · Indexed by Pub-AI: 2026-10-14 UTC
Failure Recovery via Belief Updates and World Modeling for Long-Horizon Agents
Preprint posted: 2026-10-13 · Original record ↗
- arxiv
- 2610.18082
Explicit source versions (1)
- v1 · 2026-10-13
- Preprint available
Source date: 2026-10-13 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
Long-horizon agents often struggle to recover from failures whose causes lie several steps earlier, beyond what local reflection or backtracking can resolve. We introduce BELIEFWEAVER, a framework that combines persistent evidence across attempts with prospective reasoning over repair directions. It maintains a relational belief state over frontiers, attempts, observations, and checkpoints. Execution evidence updates related frontier priorities across attempts and rollback, while diagnostic evidence gates execution. An implicit world model (IWM) evaluates repair hypotheses, predicts progress, and identifies relations among frontiers to guide exploration. We further distill these recurring IWM reasoning components into a 9B student through supervised fine-tuning and on-policy distillation (OPD). Across three terminal-based benchmarks, BELIEFWEAVER improves aggregate final score by 5.1 percentage points over native Terminus-2 continuation, with positive trends in exploration efficiency. The distilled 9B student improves final score by 9.8 percentage points over the reported native baseline while reducing estimated teacher API cost by 54.8% relative to the matched teacher. These results show that persistent evidence and prospective reasoning can improve recovery in long-horizon agent tasks while substantially reducing reliance on teacher-model inference.