π
π
The Cartographer
SD-Search: On-Policy Hindsight Self-Distillation for Search-Augmented Reasoning
May 18, 2026 Β· Grace Period Β· + Add venue
Authors
Yufei Ma, Zihan Liang, Ben Chen, Zhipeng Qian, Huangyu Dai, Lingtao Mao, Xuxin Zhang, Chenyi Lei, Wenwu Ou
arXiv ID
2605.18299
Category
cs.AI: Artificial Intelligence
Cross-listed
cs.CL,
cs.IR
Citations
0
Abstract
Search-augmented reasoning agents interleave internal reasoning with calls to an external retriever, and their performance relies on the quality of each issued query. However, under outcome-reward reinforcement learning, every search decision in a rollout shares the same trajectory-level reward, leaving individual queries without step-specific credit. Recent process-supervision approaches address this gap by drawing step-level signals from outside the policy, relying either on a much larger teacher model, or on sub-question annotations produced by a stronger external system. In contrast, we propose SD-Search, which derives step-level supervision from the policy itself through on-policy hindsight self-distillation, requiring neither an external teacher nor additional annotations. In SD-Search, a single model plays two roles that differ only in conditioning: a student that sees only the context available at inference time, and a teacher that additionally conditions on a compact hindsight block summarizing the search queries and final outcomes of a group of rollouts sampled from the same question. Since the teacher knows how each rollout unfolded and which ones succeeded, its query distribution implicitly marks which decisions were worth making, and the student is trained to recover this behavior by minimizing the token-level Jensen--Shannon divergence to the teacher at search-query positions. This layers a dense, step-level signal on top of GRPO's coarse trajectory reward. Crucially, this signal is produced by the policy itself within the standard RL training loop, without external model inference, auxiliary annotation pipeline, or additional training stage.
Community Contributions
Found the code? Know the venue? Think something is wrong? Let us know!
π Similar Papers
In the same crypt β Artificial Intelligence
R.I.P.
π»
Ghosted
Explanation in Artificial Intelligence: Insights from the Social Sciences
R.I.P.
π»
Ghosted
Federated Machine Learning: Concept and Applications
R.I.P.
π»
Ghosted
Counterfactual Explanations without Opening the Black Box: Automated Decisions and the GDPR
R.I.P.
π»
Ghosted
DeepAR: Probabilistic Forecasting with Autoregressive Recurrent Networks
R.I.P.
π»
Ghosted