NovBench (RAG) - Correctness: leaderboard

NovBench novelty assessment of EMNLP 2023 papers with a retrieval-augmented prompt with the five most similar ACL, EMNLP and NAACL 2019-2022 titles and abstracts. Correctness is the agreement between the positive, neutral and negative balance of the generated statements and that of the human reviewers, from 0 to 1, higher is better.

Metric: Correctness (DistAcc, 0-1): one minus half the L1 distance between the positive, neutral and negative proportions of the generated statements and of the human comments; retrieval-augmented prompt with the five most similar ACL, EMNLP and NAACL 2019-2022 titles and abstracts; 1,684 EMNLP 2023 papers: from the novelty-description sentences of the paper introduction (selected by GPT-5) the model writes a novelty evaluation as positive, neutral and negative statements, compared with the expert reviewers novelty comments (extracted by GPT-4o-mini, then deduplicated, consolidated and sorted by sentiment by GPT-4o); greedy decoding, at most 4,096 tokens; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 18 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1SEA-S0.674
2Qwen 3 8B0.6737#1873
3Qwen 3 32B0.6604#972
4GPT-50.6453#178
5GPT-OSS-20B0.643#1183
6DeepSeek R1 Distill Qwen 14B0.6417#2771
7DeepSeek R1 Distill Llama 70B0.626#1669
8Qwen 3 14B0.6245#1269
9GPT-OSS-120B0.6027#704
10SEA-E0.5965
11Gemini 2.5 Flash0.5923#463
12DeepReviewer-14B0.5246#844
13DeepSeek-R1-8B (NovBench checkpoint unspecified)0.522
14CycleReviewer-70B0.4892#1880
15DeepReviewer-7B0.3907

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=novbench-rag-correctness · How It Works · Data refreshed daily, snapshot 2026-11-21.