FireWorldBench - Cross-Field Coupling Understanding (Structured Records): leaderboard
Metric: Completion accuracy (%; P3, cross-field coupling understanding items of the controlled-simulation subset; structured-observation track (textual sensor traces and physical records); each cell is the mean of the matched choice-format score (Jaccard overlap of selected and gold option sets) and open-report score (exact agreement of required structured slots); deterministic evaluator, no LLM judge; failed or unparseable responses score 0). Source: arxiv.org. Saturation forecast: Around 2028. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.6 Luna | 55.49 |
| 2 | Gemma 3 27B (IT) | 46.56 |
| 3 | DeepSeek V3.2 | 45.12 |
| 4 | Kimi K2 | 44.53 |
| 5 | Qwen 3 235B A22B | 43.73 |
| 6 | Qwen 3.5 Plus (2026-02-15) | 42.76 |
| 7 | Gemini 2.5 Flash | 42.52 |
| 8 | Qwen 3 32B | 40.16 |
| 9 | Claude Haiku 4.5 | 39.92 |
| 10 | GLM 4.5 Air | 39.84 |
| 11 | Qwen 3 Next 80B A3B Instruct | 39.53 |
| 12 | Llama 3.3 70B Instruct | 37.11 |
| 13 | Qwen 3 8B | 33.88 |
| 14 | Gemma 3 12B (IT) | 33.3 |
| 15 | Llama 3.1 8B Instruct | 28.38 |
Interactive version: theaggregate.ai/benchmark?slug=fireworldbench-cross-field-coupling-understanding-structured-records · How It Works · Data refreshed daily, snapshot 2026-09-26.