Searcharxiv⌕ Search

arXiv · 2611.20034

EchoSpec: Training Search Agents to Weigh Evidence Above Echoes

Abstract

Search agents trained with reinforcement learning decide what to retrieve, when to search again, and when to answer. In real corpora, one claim is copied and paraphrased across many pages while the primary source settling a question appears once. We show that reinforcement-learned search agents grant echoes the credence of independent sources and lose their ability to abstain. Across five base models from 3B to 32B and seven released RL agents on multi-hop Wikipedia questions with primary evidence intact, eight paraphrased false-claim echoes overturn 58-73% of previously correct answers, while exact-match RL drives deleted-evidence abstention below 15%. Conditioning on retrieval localizes the failure to evidence acquisition: once primary evidence enters context, agents overrule eight echoes as reliably as one, whereas echoes fill top ranks and halt search early, so larger budgets, forced search turns, or passive passage filtering alone fail to recover displaced evidence. We introduce EchoSpec, specifying reliability across five corpus transformations with known target answers and training in two stages. Verification warm-start teaches multi-hop bridge-entity recovery past echoes, after which group RL optimizes an evidence-gated $λ$-reward with an outcome-independent acquisition bonus and scale-preserving advantages retaining the optimal rule $λ/(1+λ)$. Across Qwen2.5-3B, Qwen2.5-7B, and Llama-3.1-8B backbones on nine evaluation suites, EchoSpec cuts eight-echo deference by 30-44 points on Qwen2.5-7B down to 29-31%, nearly quadruples eight-echo reach to 26-27% with zero-shot transfer to unseen BGE and BM25 retrievers, and raises standalone QA precision to 47-55%. Code is publicly available at https://github.com/unihsy/EchoSpec.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Shihao Yang, Can Zeng, Xinzhe Li. 2026-11-15. EchoSpec: Training Search Agents to Weigh Evidence Above Echoes. https://searcharxiv.com/abs/2611.20034

Citations retain the original authors and DOI and include this site’s record URL. If this discovery service contributed to your research, you can also cite Searcharxiv by Indiciae: BibTeX · RIS.

KEEP EXPLORING

Related papers

Beyond Retrieval: Learning Compact User Representations for Scalable LLM Personalization

Personalizing large language models requires adapting model behavior to individual users while preserving robustness and deployment-scale efficiency. Existing approaches typically personalize LLMs either at the input level, by retrieving user histories or constructing profile prompts, or at the parameter level, by maintaining user-specific parameter-efficient modules. The former makes personalization sensitive to retrieval quality and prompt design, whereas the latter incurs storage and maintenance costs that grow with the user population. To address these limitations, we propose TAP-PER (Temporal Attentive Prefix for PERsonalization), a prefix-based framework that encodes user preferences as learnable representations, avoiding the serialization of user histories into prompts and replacing heavy per-user adapters with lightweight user-state prefix embeddings. Inspired by personalized recommendation systems, TAP-PER decomposes user modeling into user-state and query-conditioned components, and incorporates temporal signals to capture the evolving nature of user interests. Experiments on six LaMP tasks show that TAP-PER consistently outperforms prompt-based and model-based baselines across classification, rating, and generation settings. Moreover, TAP-PER uses 130x fewer per-user parameters than OPPU and roughly half the total parameter footprint of PER-PCS at the 1,000-user scale, demonstrating scalable personalization without prompt-serialized histories or heavy per-user adapters.

cs.IR↗

RecHarness: A Bandit-Routed Agentic Harness for Self-Evolving Recommender Systems

Optimizing recommender models remains heavily reliant on manual iteration over architectures, objectives, and training strategies. While LLM-based agents can automate this trial-and-error process, relying on LLMs to both select modification directions and formulate concrete hypotheses can lead to unstable search under limited trial budgets. Inspired by the above challenge, we propose RecHarness, a Bandit-Routed Agentic Harness for automated recommender model optimization. RecHarness decouples the optimization process into two steps: a bandit router selects the next modification direction according to historical validation feedback, while the LLM generates a concrete optimization hypothesis and executable code edit within the selected direction. To sustain long-horizon exploration, RecHarness uses a jump-basin mechanism to activate a structural-jump arm when local edits stagnate. Across multiple recommendation tasks, datasets, and model backbones, RecHarness achieves more consistent performance improvements and uses limited trial budgets more effectively than LLM-directed search. During a 7-day online A/B test on a large-scale short-video advertising platform, the selected candidate improves ADVV by 2.084%, Revenue by 0.534%, and Exposure by 0.559%. Code is available at: https://github.com/6lyc/RecHarness.

cs.IR↗

Retrieving Complementary Evidence for Multi-Hop RAG with Semantic Gain Graphs

Multi-hop retrieval-augmented generation (RAG) requires finding documents whose evidence is useful in combination. Query relevance and graph connectivity help locate related content, but neither directly measures the additional information one document contributes given another. We introduce SG$^2$RAG, a framework that represents this complementarity through semantic gain. Offline, a large language model assesses the relevant and nonredundant information contributed by one chunk given another within neighborhoods proposed by semantic similarity. These ordered assessments are aggregated into symmetric edge weights to construct a reusable Semantic Gain Graph. Online, controlled Personalized PageRank starts from query-relevant seeds and ranks complementary evidence beyond the initial retrieval set. Experiments on MuSiQue, 2Wiki, and HotpotQA with two language model backbones show that SG$^2$RAG achieves the best average EM, F1, and Recall@10 among the evaluated single-retrieval methods. Under identical candidate support and propagation, semantic gain improves average F1 and Recall@10 over similarity weighting by 4.08 and 4.31 points. Analyses of supporting documents and improvements in three iterative retrieval frameworks further support its effectiveness as a reusable retrieval component.

cs.IR↗