NovBench (Few-Shot) - Correctness: leaderboard

NovBench novelty assessment of EMNLP 2023 papers with a few-shot prompt with two analogous NovBench examples. 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; few-shot prompt with two analogous NovBench examples; 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-S0.7149
2Qwen 3 32B0.6944#983
3Qwen 3 8B0.6784#1890
4Qwen 3 14B0.6734#1284
5SEA-E0.6609
6DeepSeek R1 Distill Qwen 14B0.6606#2739
7Gemini 2.5 Flash0.659#466
8CycleReviewer-70B0.6522#1899
9DeepSeek-R1-8B (NovBench checkpoint unspecified)0.6517
10GPT-OSS-20B0.6461#1188
11DeepReviewer-7B0.6308
12DeepReviewer-14B0.6301#852
13DeepSeek R1 Distill Llama 70B0.6274#1676
14Llama-OpenReviewer-8B0.6205
15GPT-50.5411#185

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=novbench-few-shot-correctness · How It Works · Data refreshed daily, snapshot 2026-11-22.