Research map / Large language models / Retrieval and knowledge
LiveNewsBench: Evaluating Web Search Agents with Freshly Curated News
Yunfan Zhang, Kathleen McKeown, Smaranda Muresan
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add web search access (ablation) | DeepSeek V3.2 Thinking · LiveNewsBench human-verified | accuracy (%): 10.0 -> 84.5 | Improved |
| add web search access (ablation) | GPT-OSS-120B · LiveNewsBench human-verified | accuracy (%): 10.0 -> 27.0 | Improved |
| add web search access (ablation) | Gemini 3 Pro · LiveNewsBench human-verified | accuracy (%): 20.5 -> 60.5 | Improved |
| raise search budget from 1 to 7 (ablation) | DeepSeek V3.2 (No Thinking) · LiveNewsBench human-verified | accuracy (%): 20.0 -> 84.5 | Improved |
| raise search budget from 1 to 7 (ablation) | GPT-4.1 · LiveNewsBench human-verified | accuracy (%): 62.5 -> 66.5 | Improved |
| raise search budget from 1 to 7 (ablation) | Claude Sonnet 4.5 · LiveNewsBench human-verified | accuracy (%): 53.5 -> 67.0 | Improved |
| raise search budget from 5 to 7 (ablation) | Claude Sonnet 4.5 · LiveNewsBench human-verified | accuracy (%): 82.0 -> 67.0 | Hurt |
| raise search budget from 5 to 7 (ablation) | Gemini 3 Pro · LiveNewsBench human-verified | accuracy (%): 60.5 -> 74.5 | Improved |
| extend tool budget to 15 searches and 15 visits (ablation) | DeepSeek V3.2 Thinking · LiveNewsBench human-verified | accuracy (%): 84.5 -> 84.0 | No clear effect |
| extend tool budget to 15 searches and 15 visits (ablation) | Claude Sonnet 4.5 · LiveNewsBench human-verified | accuracy (%): 82.0 -> 81.0 | No clear effect |
| use official web search API instead of local agent framework | GPT-5.2 · LiveNewsBench human-verified | accuracy (%): 74.0 -> 90.0 | Improved |
| use official web search API instead of local agent framework | Claude Sonnet 4.5 · LiveNewsBench human-verified | accuracy (%): 82.0 -> 40.0 | Hurt |
Closest papers by meaning
- Mind2Web 2: Evaluating Agentic Search with Agent-as-a-Judge (NeurIPS 2025 · similarity 0.90)
- WideSearch: Benchmarking Agentic Broad Info-Seeking (ICLR 2026 · similarity 0.90)
- AgentWebBench: Benchmarking Multi-Agent Coordination in Agentic Web (ICML 2026 · similarity 0.88)
- Iterative Self-Incentivization Empowers Large Language Models as Agentic Searchers (NeurIPS 2025 · similarity 0.88)
- Empowering Efficiency and Efficacy in WebAgent via Enabling Info-Rich Seeking (ICLR 2026 · similarity 0.87)
- AgentHop: A Diagnostic Benchmark for Agentic Multi-Hop Scientific Question Answering (NeurIPS 2026 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
