paper-with-me

홈 › Papers

Exploring Information Seeking Agent Consolidation

2026-01-31 · Guochen Yan, Jialong Wu, Zhengwei Tao, Bo Li, Qintong Zhang, Jiahao Xu, Haitao Mi, Yuejian Fang, Qingni Shen, Wentao Zhang, Zhonghai Wu arxiv

Information-seeking agents have emerged as a powerful paradigm for knowledge-intensive tasks, yet today's systems remain specialized for the open web, documents, or local knowledge bases, hindering scalable and cross-domain deployment. We present the first systematic empirical study of consolidating these information-seeking agents into a single foundation agentic model. We compare two paradigms -- \emph{data-level mixing}, which trains a unified model on a mixture of datasets, and \emph{parameter-level merging}, which merges independently trained experts in parameter space -- across 3 training scenarios, evaluating \textbf{26} representative parameter-level methods on \textbf{10} benchmarks. To compare across heterogeneous benchmarks, we introduce a geometric Composite Score and an Imbalance Score that describe overall performance and task skew. Our analysis shows that (i) well-designed parameter-level merging attains parity with data mixing at a fraction of its training cost and is order-agnostic; (ii) parameter-level merging structurally preserves out-of-domain capabilities that data mixing universally forgets; and (iii) cross-scenario stability is strongly tied to consolidation quality. We distil our observations into a method-selection guide and design principles for next-generation merging operators.

📄 PDF Abstract BibTeX arXiv:2602.00585

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

AgentFold: Long-Horizon Web Agents with Proactive Context Management

2025-10-28 · Rui Ye, Zhongwang Zhang, Kuan Li, Huifeng Yin 외 arxiv

LLM-based web agents show immense promise for information seeking, yet their effectiveness on long-horizon tasks is hindered by a fundamental trade-off in context management. Prevailing ReAct-based agents suffer from con…

WideSeek-R1: Exploring Width Scaling for Broad Information Seeking via Multi-Agent Reinforcement Learning

2026-02-04 · Zelai Xu, Zhexuan Xu, Ruize Zhang, Chunyang Zhu 외 arxiv

Recent advancements in Large Language Models (LLMs) have largely focused on depth scaling, where a single agent solves long-horizon problems with multi-turn reasoning and tool use. However, as tasks grow broader, the key…

Multi-agent Reinforcement Learning

Exploration via Empowerment Gain: Combining Novelty, Surprise and Learning Progress

2021-06-13 · ICML Workshop URL 2021 7 · Philip Becker-Ehmck, Maximilian Karl, Jan Peters, Patrick van der Smagt

Exploration in the absence of a concrete task is a key characteristic of autonomous agents and vital for the emergence of intelligent behaviour. Various intrinsic motivation frameworks have been suggested, such as novelt…

Episodic-to-Semantic Consolidation Without Identity Drift

2026-07-02 · Xue Qin, Simin Luan, Cong Yang, Zhijun Li arxiv

Long-running adaptive intelligent agents face a structural tension between knowledge consolidation and information integrity. Memory consolidation is conventionally treated as an agent-changing operation: a model is fine…

InfoDeepSeek: Benchmarking Agentic Information Seeking for Retrieval-Augmented Generation

2025-05-21 · Yunjia Xi, Jianghao Lin, Menghui Zhu, Yongzhao Xiao 외

Retrieval-Augmented Generation (RAG) enhances large language models (LLMs) by grounding responses with retrieved information. As an emerging paradigm, Agentic RAG further enhances this process by introducing autonomous L…

BenchmarkingRAGRetrievalRetrieval-augmented Generation