CoSearchAgent: A Lightweight Collaborative Search Agent with Large Language Models
Authors: Peiyuan Gong, Jiamian Li, Jiaxin Mao Published: 2024-02-09 Source: https://arxiv.org/abs/2402.06360 Type: Demo paper (Conversational Search)
Overview
Collaborative search supports multiple users working together to accomplish a specific search task. Prior research found that lightweight collaborative search plugins inside instant messaging platforms align well with users' collaborative habits. However, the complexity of multi-user interaction made fully functional systems difficult to build, so earlier studies relied on the Wizard of Oz paradigm. Leveraging the ability of LLM-based agents to interact naturally with users and complete complex information-seeking tasks, the authors propose CoSearchAgent, a lightweight collaborative search agent powered by LLMs.
Key capabilities
- Designed as a Slack plugin supporting collaborative search during multi-party conversations.
- Understands queries and conversational context across multiple users.
- Searches the Web for relevant information via APIs and answers queries grounded in the retrieved results.
- Can ask clarifying questions when user information needs are unclear.
- Highly flexible; intended as a research platform for collaborative search.
- Code and demo video are publicly accessible.
- Classical cascades (retrieval → reranking → generation) remain dominant, but agentic paradigms make retrieval strategies themselves learnable.
- High-quality instruction data and session logs matter as much as model architecture; synthetic data risks leakage and distribution shift.
- Offline metrics increasingly diverge from online satisfaction; LLM-as-judge should be cross-validated with human evaluation.
- Latency, cost, interpretability, and safety are hard constraints for industrial deployment.
- A Survey of Conversational Search (ACM, Sep 2025)
- Engineering Conversational Search Systems (arXiv:2407.00997)
- Agentic Conversational Search with Contextualized Reasoning
- CTR-Guided Generative Query Suggestion in Conversational Search (EMNLP)
- ChatRetriever: Adapting LLMs for Generalized Retrieval (arXiv:2404.13556)
- ConvGQR: Generative Query Reformulation for Conversational Search (arXiv:2305.15645)
- Original paper: *CoSearchAgent: A Lightweight Collaborative Search Agent with Large Language Models* — arXiv:2402.06360
Context within search and recommendation research
The forum post situates CoSearchAgent in the intersection of agentic search and large-scale search/personalization systems. It notes broader trends:
Engineering checklist (from the post)
| Item | Concern | Suggestion | |------|---------|------------| | Data | PII, index versioning | Partitioned indexes, anonymization, rollback-capable embeddings | | Latency | p99 budget, retrieval steps | Cascades with early stopping, query caching, async reranking | | Quality | Offline vs. online gains | Interleaving experiments, human audits, citation checks | | Safety | Poisoning/bias from open retrieval | Source allowlists, adversarial detection, output filtering | | Cost | Token and GPU cost per query | Model routing, distillation, hybrid sparse-dense retrieval |