arXiv 2610.05114 · cs.CL · 2026-10-04
Belief-Trajectory Energy: Measuring the Path to a Prediction
Jiahao Ying, Wei Tang, Boxian Ai, Yaoning Wang, Haotian Chen, Wenhe Sun, Caijun Xu, Haozhan Cai 等 10 人
机构待核实
在最新榜 #391 本站收录于 2026-10-25
一句话要点
BTE（信念轨迹能量）衡量大模型在各层逐步修正预测的过程，而不只看最终输出，理论上对应 Fisher-Rao 几何下的预测修正，可作为推理任务难度信号。
Belief-Trajectory Energy measures how an LLM revises its predictions layer by layer rather than only the final output, grounded in Fisher-Rao geometry, and serves as a model-relative difficulty signal on reasoning tasks.
阅读 arXiv 原文 PDF Hugging Face 讨论 项目页
主题与学科
同主题：大语言模型
- Jeva: System One Models from First Principles ▲371
- Post-Training Leaves Behavioral Shadows on Unrelated Decisions ▲273
- OmniEdu: Open Foundation Models for Learning and Teaching ▲240
- PAR: Progress-Aligned Recurrence for Length Generalization in State Tracking ▲163
- MemT-1.0: Towards Memory-Native Evolution ▲143
同主题：推理
- LoopVL: Recurrent Visual Intelligence 2026-09-29 · ▲417
- LEV: Thinking Fast and Deep with Looped Decision Models 2026-10-14 · ▲337
- Thinking Inertia: LLMs Keep Thinking When Told Not To 2026-10-08 · ▲241
- Rethinking Latent Visual Reasoning: Grounding Latent Reasoning in Visual Evidence 2026-09-28 · ▲224
- On-Policy or Off-Policy Learning? A Systematic Study of Distillation Dynamics 2026-09-28 · ▲205
数据说明
元数据来自 arXiv API（已核实）；点赞来自 Hugging Face 每日论文；引用来自 Semantic Scholar。最近刷新 2026-10-26。
要点由本站根据摘要自写，不转载摘要原文，请阅读原文。