BacktestBench - Indicator Retrieval: leaderboard
Metric: Accuracy (%) of the Summarizer stage in retrieving the factors and indicators a strategy uses (BM25-assisted), synthetic BacktestBench test split (the expert-crafted subset excluded), each model driving the AutoBacktest pipeline of the authors (Summarizer, SQL Retriever and Python Coder agents) over a PostgreSQL market database, temperature 0.6; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 23 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro | 87.15 |
| 2 | Kimi K2 (Thinking) | 79.82 |
| 3 | Seed 1.8 | 76.93 |
| 4 | Qwen 3 235B A22B 2507 (Thinking) | 76.41 |
| 5 | Kimi Linear 48B A3B Instruct | 74.37 |
| 6 | GLM-4.7 | 74.35 |
| 7 | Qwen 3 Next 80B A3B (Thinking) | 73.8 |
| 8 | Qwen 3 Max | 72.42 |
| 9 | Qwen 3 30B A3B 2507 (Thinking) | 69.73 |
| 10 | DeepSeek V3.2 | 68.98 |
| 11 | GPT-OSS-120B | 67.65 |
| 12 | GPT-OSS-20B | 67.54 |
| 13 | Qwen 3 32B | 67.41 |
| 14 | Qwen 3 Coder Plus | 65.09 |
| 15 | Qwen 3 14B | 60.92 |
Interactive version: theaggregate.ai/benchmark?slug=backtestbench-indicator-retrieval · How It Works · Data refreshed daily, snapshot 2026-10-07.