Same Trajectory, Different Knowledge: Information-State-Aware On-Policy Distillation for Search Agents
cs.IR
Submitted: 2026-11-01
Updated: 2026-11-01
Terminology
Sources
- Rethinking Uncertainty Estimation in LLMs: A Principled Single-Sequence Measure
- Know More, Know Clearer: A Meta-Cognitive Framework for Knowledge Augmentation in Large Language Models
- Better than Your Teacher: LLM Agents that learn from Privileged AI Feedback
- KbSD: Knowledge Boundary aware Self-Distillation for Behavioral Calibration in Agentic Search
- PiCA: Pivot-Based Credit Assignment for Search Agentic Reinforcement Learning
- Self-Distilled Agentic Reinforcement Learning
- SD-Search: On-Policy Hindsight Self-Distillation for Search-Augmented Reasoning
- Privileged Information Distillation for Language Models
- Qwen2.5 Technical Report
- Proximal Policy Optimization Algorithms
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- From Self-Distillation to Self-Practice: Privileged Information for Multi-Turn Agents
- GAPD: Gold-Action Policy Distillation for Agentic Reinforcement Learning in Knowledge Base Question Answering
- RACE: Learning Multimodal Search through Referential Abstraction and Compositional Evidence
- Skill-SD: Skill-Conditioned Self-Distillation for Multi-turn LLM Agents
- EviSD: Evidence-Conditioned Self-Distillation for Search-Augmented Agents
- Self-Distilled RLVR
- DOPD: Dual On-policy Distillation
- Beyond Absolute Imitation: Anchored Residual Guidance for Privileged On-Policy Distillation
Related papers
- UniRank: Benchmarking Ranking Models for Unified Sequential Modeling and Feature Interaction
- Revisiting Text Ranking in Deep Research
- GateSID: One Gate to Balance Semantic and Collaborative Signals across the Item Lifecycle
- SAM-D2Q: Aligning Multimodal Doc2Query with Search Demand and Conversion for E-commerce
- GINGER: Spicing Up Graph-Based Vector Search
- Subtraction Gets You More: Gap-Aware Retrieval for Multimodal Multi-Hop QA