TKFQA - Reasoning-Chain Accuracy (Table-Text-KG Order): leaderboard
TKFQA (2026) asks counterfactual multi-hop questions whose evidence is spread over a table, text passages and a knowledge graph. Reasoning-chain accuracy is the entity-level agreement of the generated chain with the annotated counterfactual chain, in percent; higher is better, with the contexts given in table, text, knowledge-graph order.
Metric: Reasoning-chain accuracy (%; entity-level agreement of the generated reasoning chain with the annotated counterfactual chain, for counterfactual multi-hop questions grounded in a table, text passages and a knowledge graph, 1,012 test items; contexts given in table, text, knowledge-graph order; mean of three runs). Source: arxiv.org. Saturation forecast: Around February 2027. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O3 | 79.05 | #224 |
| 2 | Gemini 2.5 Flash | 77.93 | #452 |
| 3 | DeepSeek V3.2 | 69.2 | #403 |
| 4 | O4 Mini | 68.18 | #354 |
| 5 | DeepSeek V4 Flash | 65.26 | #254 |
| 6 | Kimi K2.5 | 64.04 | #252 |
| 7 | MiniMax-M2.7 | 55.34 | #519 |
| 8 | GPT-5 | 51.52 | #163 |
| 9 | Qwen 3 30B A3B | 43.31 | #1124 |
| 10 | Gemini 2.5 Flash Lite | 37.25 | #958 |
| 11 | Grok 4.3 | 32.61 | #332 |
| 12 | GPT-4.1 Mini | 29.59 | #775 |
| 13 | Qwen 3 8B | 20.14 | #1838 |
| 14 | Gemma 3 12B | 19.95 | #2013 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=tkfqa-reasoning-chain-accuracy-table-text-kg-order · How It Works · Data refreshed daily, snapshot 2026-11-12.