Research Papers 论文研究 22h ago Updated 1h ago 更新于 1小时前 50

Teaching LLMs to Update Beliefs for Efficient Long-Horizon Interaction 教授大语言模型更新信念以实现高效长周期交互

ABBEL introduces a framework that isolates and supervises summary generation in long-context LLM interactions by treating summaries as belief states, reducing performance degradation compared to traditional recursive summarization. Belief grading, inspired by autoencoding, supervises belief states by rewarding the model’s ability to reconstruct recent observations from the belief, improving learning efficiency and reducing training steps by 50%. ABBEL reduces the performance gap to full-context ABBEL 提出将自然语言信念状态(belief states)作为摘要,替代传统递归摘要,通过信念分级(belief grading)监督信息内容,解决长交互任务中上下文扩展受限问题。 引入自编码启发式的信念分级机制,使模型在训练时仅需关注关键信息保留,显著降低训练步数并缩小与全上下文模型的性能差距。 在协作代码生成基准 CollabBench 上,ABBEL 将性能差距缩小约 50%,训练步数减少 50%,同时峰值上下文 token 数量大幅降低。 该方法有效缓解了人类辅助场景中高质量训练数据稀缺导致的自摘要性能退化问题,提升了模型在真实交互环境中的泛化能力。 ABBEL 为长上下文智能体

70
Hot 热度
75
Quality 质量
68
Impact 影响力

Analysis 深度分析

TL;DR

  • ABBEL introduces a framework that isolates and supervises summary generation in long-context LLM interactions by treating summaries as belief states, reducing performance degradation compared to traditional recursive summarization.
  • Belief grading, inspired by autoencoding, supervises belief states by rewarding the model’s ability to reconstruct recent observations from the belief, improving learning efficiency and reducing training steps by 50%.
  • ABBEL reduces the performance gap to full-context models by ~50% in collaborative coding tasks (CollabBench) while using significantly fewer peak context tokens, enabling scalable, memory-efficient long-horizon assistance.

Why It Matters

This work addresses a critical bottleneck in deploying LLMs for complex, long-horizon tasks like collaborative software development, where context length is constrained and human feedback is scarce. By decoupling summarization from task execution and supervising it via belief grading, ABBEL offers a practical path to maintain high performance without requiring massive context windows or perfect training data—key for real-world assistive AI systems.

Technical Details

  • ABBEL reformulates context summarization as belief state updates, drawing from recursive Bayesian estimation, where each belief state $b_{t+1}$ is a compressed, interpretable representation of the interaction history up to step $t$.
  • Belief grading uses an auxiliary RL objective: the model is rewarded for its ability to reconstruct the most recent observation $o_t$ from the updated belief $b_{t+1}$, using the current policy $\pi_\theta$ as both encoder and decoder—effectively treating the belief as a latent code.
  • The reconstruction grading objective is defined as: $\mathcal{L}{\text{grade}} = \log P(o_t \mid b{t+1}, a_t, b_t, p_I)$, where $p_I$ is the task prompt, ensuring beliefs retain information critical for decoding recent context.
  • Evaluated on CollabBench, a human-assisted coding benchmark, ABBEL with reconstruction grading (ABBEL-rec-BG) achieves 0.48 test pass rate (vs. 0.52 for full context) using only ~600 peak tokens (vs. 1408), while training in half the steps compared to non-graded summarization.

Industry Insight

  • For AI developers building long-context assistants (e.g., coding copilots, customer support agents), ABBEL provides a plug-in architecture to maintain performance under context limits without retraining from scratch—ideal for systems with constrained memory or high interaction latency.
  • The belief grading framework is domain-agnostic and can be adapted to other assistive tasks (e.g., legal document review, medical triage) by designing reconstruction rewards that reflect task-specific fidelity, enabling scalable, interpretable summarization in low-data regimes.

TL;DR

  • ABBEL 提出将自然语言信念状态(belief states)作为摘要,替代传统递归摘要,通过信念分级(belief grading)监督信息内容,解决长交互任务中上下文扩展受限问题。
  • 引入自编码启发式的信念分级机制,使模型在训练时仅需关注关键信息保留,显著降低训练步数并缩小与全上下文模型的性能差距。
  • 在协作代码生成基准 CollabBench 上,ABBEL 将性能差距缩小约 50%,训练步数减少 50%,同时峰值上下文 token 数量大幅降低。
  • 该方法有效缓解了人类辅助场景中高质量训练数据稀缺导致的自摘要性能退化问题,提升了模型在真实交互环境中的泛化能力。
  • ABBEL 为长上下文智能体提供了一种可解释、高效且低内存占用的上下文管理范式,适用于代码生成、对话系统等需要多轮交互的领域。

为什么值得看

随着任务复杂度提升,LLM 的上下文窗口面临物理与计算极限,传统递归摘要虽能压缩历史但严重损害性能,尤其在人机协作等数据稀缺场景。ABBEL 通过信念状态建模与分级监督,提供了一种兼顾效率与性能的新路径,对构建可持续演进的智能体系统具有关键指导意义。

技术解析

  • ABBEL 将对话或任务历史中的关键信息提炼为“信念状态”(belief state),作为模型的当前工作上下文,取代完整交互历史,实现上下文压缩与语义聚焦。
  • 信念更新机制借鉴递归贝叶斯估计,模型在每个时间步根据 prior belief、动作和观测生成 posterior belief,形成交替更新与决策的闭环流程。
  • 提出“信念分级”作为辅助强化学习目标:利用当前模型 πθ 作为解码器,尝试从新信念 bt+1 重构最新观测 ot,重构质量越高则奖励越大,从而引导信念保留关键信息。
  • 在缺乏领域特定启发式时,采用通用自编码式重建目标(Eq. 1),使信念成为历史信息的压缩编码,兼具可解释性与训练稳定性。
  • 实验基于 CollabBench 协作代码环境,测试模型在有限训练步数下完成函数生成与单元测试的能力,对比全上下文、无分级摘要及 ABBEL-rec-BG 三种设置。

行业启示

  • 面向长交互任务(如编程助手、客服机器人),应优先采用结构化信念管理而非简单截断或递归摘要,以平衡上下文长度与语义完整性。
  • 在数据稀缺的人机协作场景中,引入自监督的信念分级机制可显著提升样本效率,减少对大规模人类轨迹数据的依赖,降低部署成本。
  • 未来智能体架构设计应将“信念压缩”视为核心模块,结合领域知识定制分级奖励函数,同时保留通用重建目标以增强跨任务适应性。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

LLM 大模型 Agent Agent Conversational AI 对话系统 Training 训练 Evaluation 评测