ICA: Information-Aware Credit Assignment for Visually Grounded Long-Horizon Information-Seeking Agents
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866912897293090816 |
|---|---|
| author | Pang, Cong Feng, Xuyu Yi, Yujie Chen, Zixuan Hong, Jiawei Yao, Tiankuo Yuan, Nang Luo, Jiapeng Lu, Lewei Lou, Xin |
| author_facet | Pang, Cong Feng, Xuyu Yi, Yujie Chen, Zixuan Hong, Jiawei Yao, Tiankuo Yuan, Nang Luo, Jiapeng Lu, Lewei Lou, Xin |
| contents | Despite the strong performance achieved by reinforcement learning-trained information-seeking agents, learning in open-ended web environments remains severely constrained by low signal-to-noise feedback. Text-based parsers often discard layout semantics and introduce unstructured noise, while long-horizon training typically relies on sparse outcome rewards that obscure which retrieval actions actually matter. We propose a visual-native search framework that represents webpages as visual snapshots, allowing agents to leverage layout cues to quickly localize salient evidence and suppress distractors. To learn effectively from these high-dimensional observations, we introduce Information-Aware Credit Assignment (ICA), a post-hoc method that estimates each retrieved snapshot's contribution to the final outcome via posterior analysis and propagates dense learning signals back to key search turns. Integrated with a GRPO-based training pipeline, our approach consistently outperforms text-based baselines on diverse information-seeking benchmarks, providing evidence that visual snapshot grounding with information-level credit assignment alleviates the credit-assignment bottleneck in open-ended web environments. The code and datasets will be released in https://github.com/pc-inno/ICA_MM_deepsearch.git. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_10863 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | ICA: Information-Aware Credit Assignment for Visually Grounded Long-Horizon Information-Seeking Agents Pang, Cong Feng, Xuyu Yi, Yujie Chen, Zixuan Hong, Jiawei Yao, Tiankuo Yuan, Nang Luo, Jiapeng Lu, Lewei Lou, Xin Machine Learning Artificial Intelligence Despite the strong performance achieved by reinforcement learning-trained information-seeking agents, learning in open-ended web environments remains severely constrained by low signal-to-noise feedback. Text-based parsers often discard layout semantics and introduce unstructured noise, while long-horizon training typically relies on sparse outcome rewards that obscure which retrieval actions actually matter. We propose a visual-native search framework that represents webpages as visual snapshots, allowing agents to leverage layout cues to quickly localize salient evidence and suppress distractors. To learn effectively from these high-dimensional observations, we introduce Information-Aware Credit Assignment (ICA), a post-hoc method that estimates each retrieved snapshot's contribution to the final outcome via posterior analysis and propagates dense learning signals back to key search turns. Integrated with a GRPO-based training pipeline, our approach consistently outperforms text-based baselines on diverse information-seeking benchmarks, providing evidence that visual snapshot grounding with information-level credit assignment alleviates the credit-assignment bottleneck in open-ended web environments. The code and datasets will be released in https://github.com/pc-inno/ICA_MM_deepsearch.git. |
| title | ICA: Information-Aware Credit Assignment for Visually Grounded Long-Horizon Information-Seeking Agents |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2602.10863 |