EditJudge-Bench - Counterfactual - Sem.: leaderboard
EditJudge-Bench v1,1382curatedBlender before/after pairs,factual/counterfactual instructions. NativeAUROC or multiplechoice factual-instruction ranking percent. No parsing-failure dropping:asymmetric score2fail factual,8failcounterfactual;unsupportedlogitoperation.5;APIerrors score5;zero outputs retained. Exact publishedTOML backend/model/prompt/image-order/budget retained; source-unreported launcherflags remain unknown. Gemma31B is NVIDIA NVFP4;QwenThinking is its distinct checkpoint. Supervised judges retain own trainedartifact IDs,not their basemodels.
Metric: AUROC (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 25 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | gpt-5.6-sol (EditJudge Run: config=vie__gpt-5.6-sol;backend=openai,template=vie_tie_0shot,image_position=after,reasoning_effort=low,max_tokens=8192) | 89.5 | |
| 2 | claude-opus-5 (EditJudge Run: config=vie__claude-opus-5;backend=anthropic,template=vie_tie_0shot,image_position=after,thinking_effort=low,max_tokens=8192) | 89.3 | |
| 3 | gpt-5.4 (EditJudge Run: config=vie__gpt-5.4;backend=openai,template=vie_tie_0shot,image_position=after,reasoning_effort=low,max_tokens=8192) | 84.8 | |
| 4 | Gemma-4-31B-IT-NVFP4 (EditJudge Run: config=logprob__gemma-4-31b;backend=vllm,template=binary_q_images_last,logprobs=True,max_model_len=4096) | 83 | |
| 5 | gemini-2.5-flash (EditJudge Run: config=vie__gemini-2.5-flash;backend=gemini,template=vie_tie_0shot,image_position=after,thinking=True,max_tokens=128) | 80.6 | |
| 6 | Gemma-4-31B-IT-NVFP4 (EditJudge Run: config=vie__gemma-4-31b;backend=vllm,template=vie_tie_0shot,image_position=after,max_tokens=16,max_model_len=4096) | 80 | |
| 7 | gpt-4o (EditJudge Run: config=vie__gpt-4o;backend=openai,template=vie_tie_0shot,image_position=after,max_tokens=1024) | 77.9 | |
| 8 | Gemma-4-31B-IT-NVFP4 (EditJudge Run: config=kris__gemma-4-31b;backend=vllm,template=kris_instruction,image_position=after,max_tokens=16,max_model_len=4096) | 77.7 | |
| 9 | Gemma-4-31B-IT-NVFP4 (EditJudge Run: config=imgedit__gemma-4-31b;backend=vllm,template=imgedit,image_position=after,max_tokens=16,max_model_len=4096;unlogged_runtime_fields=max_tokens,image_position;values=reviewed_config_defaults_not_confirmed_launch) | 77.3 | |
| 10 | Qwen3-VL-32B-Instruct (EditJudge Run: config=vie__qwen3-vl-32b;backend=vllm,template=vie_tie_0shot,image_position=after,max_tokens=16,max_model_len=4096) | 75.3 |
Interactive version: theaggregate.ai/benchmark?slug=editjudge-bench-counterfactual-sem · How It Works · Data refreshed daily, snapshot 2026-11-12.