Problem
Research questions and friction points this paper is trying to address.
trajectory data
large language models
open-ended tasks
ground truth
annotation cost
From Final Artifacts to Trajectories: Retrospective Process Supervision for Evidence-Grounded Long-Form Generation
🤖 AI Summary
本文提出RetroGen框架，通过从高质量最终成果中重构潜在轨迹并验证，以解决开放任务中缺乏轨迹数据的问题，提升模型的证据导向生成能力。
📝 Abstract
Trajectory data is getting more vital for training large language models for boosting the agentic abilities. Unlike the verifiable domains such as coding or mathematics, scaling trajectory data for open-ended tasks is much more difficult because these tasks lack singular ground truth and are costly to annotate or verify. In this paper, we propose RetroGen, a self-improving framework of retrospective process supervision. Our key observation is that although expert trajectories are scarce, high-quality final artifacts such as literature reviews, analyst reports and legal judgments, are abundant in pre-training data and can be viewed as compressed traces of the evidence-seeking processes that produced them. RetroGen reconstructs candidate latent trajectories from expert artifacts, verifies them against both the artifact and supporting evidence, and trains models on their own successful reconstruction data, without requiring trajectory data from stronger models. Experiments show that RetroGen improves grounding, faithful synthesis, and long-form evidence-seeking agent tasks.
Innovation
Methods, ideas, or system contributions that make the work stand out.
RetroGen
Retrospective Process Supervision
Latent Trajectories
Evidence-Grounded Long-Form Generation
Self-Improving Framework
🔎 Similar Papers
No similar papers found.
