CDR-Bench - Order-Agnostic Recipes: leaderboard
CDR-Bench (2026) tests whether LLMs faithfully execute data-refinement recipes written in natural language over web, LaTeX, RAG-preparation and privacy text, using 29 Data-Juicer operators with deterministic reference outputs. This board is the percent of order-agnostic mapper recipes (several rewriting operators whose order does not matter) that the model gets exactly right within three attempts, with each model in non-thinking mode. Higher is better.
Metric: Recipe success RS@3 (%) on the order-agnostic mapper-recipe tasks of CDR-Bench (3,462 data-refinement tasks over web, LaTeX, RAG-preparation and privacy text with 29 Data-Juicer operators and deterministic reference outputs): a task counts when the model's refined text or KEEP/DROP judgment exactly matches the reference within 3 attempts; non-thinking mode, baseline prompt; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 4 31B (IT) (Non-reasoning) | 25.55 | #342 (Gemma 4 31B (IT)) |
| 2 | GLM-5 (Non-reasoning) | 24.8 | #260 (GLM-5) |
| 3 | Kimi K2.6 (Non-reasoning) | 24.14 | #178 (Kimi K2.6) |
| 4 | DeepSeek V4 Flash (Non-reasoning) | 23.94 | #258 (DeepSeek V4 Flash) |
| 5 | Qwen 3.6 Max | 23.94 | #180 |
| 6 | Qwen 3.6 27B (Non-reasoning) | 23.73 | #341 (Qwen 3.6 27B) |
| 7 | GPT-5.4 (Non-reasoning) | 22.74 | #123 (GPT-5.4) |
| 8 | Claude Opus 4.6 | 22.33 | #102 |
| 9 | Claude Opus 4.5 | 22.13 | #156 |
| 10 | Qwen 3.6 35B A3B (Non-reasoning) | 20.4 | #462 (Qwen 3.6 35B A3B) |
| 11 | Llama 4 Scout | 13.59 | #1776 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=cdr-bench-order-agnostic-recipes · How It Works · Data refreshed daily, snapshot 2026-11-13.