HoosierHelp - Resource Exact Match: leaderboard
HoosierHelp (Li et al., 2026) tests LLM agents that help simulated users find Indiana public social-service resources through a structured search tool, across single and composite needs, unsatisfiable constraints and difficult user behaviours. This board is the percentage of 240 conversations whose final recommendation exactly matches the target resource set using only resources returned by earlier searches; higher is better.
Metric: Resource exact match (%; conversations ending in a final_recommendation whose resource IDs, all returned by an earlier search, equal the target set; 240 simulated social-service conversations over 3,971 Indiana public resources, balanced over single and composite needs, direct-match and fallback-required constraints and five simulated-user behaviours; GPT-4.1-mini simulated users, at most eight interaction turns). Source: arxiv.org. Saturation forecast: Around February 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.7 Max | 67.5 | #145 |
| 2 | GPT-5.4 | 64.17 | #123 |
| 3 | Qwen 3.6 35B A3B | 61.67 | #462 |
| 4 | DeepSeek V4 Flash | 57.5 | #258 |
| 5 | GPT-4.1 Mini | 57.08 | #779 |
| 6 | Qwen 3.6 27B | 57.08 | #341 |
| 7 | GPT-OSS-120B | 45.42 | #700 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=hoosierhelp-resource-exact-match · How It Works · Data refreshed daily, snapshot 2026-11-13.