Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note2 min read

ABSeeker: Training Long-Horizon Search Agents via Answer-Backtracked Credit Assignment

Source

arXiv

arxiv.org/abs/2608.05102v1

Read the source

Source

Plain-Language Brief

The paper proposes a new method called Answer-Backtracked Credit Assignment (ABC) for training long-horizon search agents that make multiple sequential actions to reach an answer. ABC focuses on rewarding useful actions and suppressing erroneous or redundant ones by converting sparse trajectory-level outcomes into dense step-level supervision. The method is applied in the training of ABSeeker, which achieves better results than same-scale agents and even competes with larger ones.

What the Document Says

  • Long-horizon search agents make multiple sequential actions to reach an answer.
  • Existing methods for training these agents treat all steps uniformly during supervised fine-tuning (SFT) and reinforcement learning (RL).
  • The paper proposes Answer-Backtracked Credit Assignment (ABC) to convert sparse trajectory-level outcomes into dense step-level supervision.
  • ABC rewards useful actions while suppressing erroneous or redundant actions.
  • ABSeeker is trained using ABC, achieving better results than same-scale agents.

Key Concepts Mentioned

  • Long-horizon search agents
  • Answer-Backtracked Credit Assignment (ABC)
  • Supervised fine-tuning (SFT)
  • Reinforcement learning (RL)
  • Qwen3.5-4B

Claims Made by the Document

  • ABC can effectively train long-horizon search agents.
  • ABSeeker outperforms same-scale agents and competes with larger ones.

Evidence Used

  • The paper describes the method and its application in training ABSeeker.
  • Results show that ABSeeker achieves better performance than other agents.

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Long-horizon search agents must make multiple sequential actions (steps) to search, retrieve, verify, and integrate evidence to reach a final answer. However, existing methods for training these agents typically treat all steps within a trajectory uniformly during both supervised fine-tuning (SFT) and reinforcement learning (RL), failing to distinguish useful actions from erroneous or redundant ones. In this paper, we propose Answer-Backtracked Credit Assignment (ABC), a fine-grained credit assignment framework for training long-horizon search agents by converting sparse trajectory-level outcomes into dense step-level supervision that rewards useful actions (even in failed trajectories) while suppressing erroneous or redundant actions. Specifically, given a potentially obscure query and its corresponding ground-truth answer, ABC first performs Answer-Backtracked Clue Recovery, which traces back from the answer to recover intermediate clues required to solve the question. It then applies Clue-Anchored Step Scoring to evaluate each search step against these clues, converting sparse binary outcome supervision into dense step-level rewards. Based on these rewards, we develop ABC-SFT, which reweights the loss of each turn, and ABC-GRPO, which uses the step-level scores as rewards in GRPO. Building on this framework, we train ABSeeker based on Qwen3.5-4B with only 8.5k examples. ABSeeker achieves 37.3% on BrowseComp and 39.1% on BrowseComp-ZH. With context management, the scores further improve to 55.3% and 52.9%, respectively, significantly outperforming same-scale (4B) agents and even matching the performance of larger ones (approximately 30B). These results demonstrate the effectiveness of answer-backtracked step-level credit assignment for training long-horizon search agents.