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

Researchers propose a method called Answer-Backtracked Credit Assignment (ABC) to train long-horizon search agents more effectively. This method converts sparse trajectory-level outcomes into dense step-level supervision, rewarding useful actions and suppressing erroneous or redundant ones. The trained agent, named ABSeeker, achieves impressive results on two benchmarks, outperforming other agents of similar size and even competing with larger ones.

What the Document Says

  • The paper presents a method called Answer-Backtracked Credit Assignment (ABC) for training long-horizon search agents.
  • ABC converts sparse trajectory-level outcomes into dense step-level supervision, rewarding useful actions and suppressing erroneous or redundant ones.
  • The method involves Answer-Backtracked Clue Recovery and Clue-Anchored Step Scoring.
  • Based on this framework, ABSeeker was trained with only 8.5k examples from Qwen3.5-4B.
  • ABSeeker achieved 37.3% on BrowseComp and 39.1% on BrowseComp-ZH.
  • With context management, the scores further improved to 55.3% and 52.9%, respectively.

Key Concepts Mentioned

  • long-horizon search agents
  • Answer-Backtracked Credit Assignment (ABC)
  • Qwen3.5-4B
  • BrowseComp
  • BrowseComp-ZH

Claims Made by the Document

  • ABC is an effective method for training long-horizon search agents.
  • ABSeeker outperforms other agents of similar size and even competes with larger ones.

Evidence Used

  • The results demonstrate the effectiveness of answer-backtracked step-level credit assignment for training long-horizon search agents.
  • ABSeeker achieved impressive scores on two benchmarks.

Methods / Approach

  • Answer-Backtracked Clue Recovery
  • Clue-Anchored Step Scoring
  • ABC-SFT
  • ABC-GRPO

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.