Codesota · Natural Language Processing · Natural Language Inference · SNLITasks/Natural Language Processing/Natural Language Inference
Natural Language Inference · benchmark dataset · 2015 · EN

Stanford Natural Language Inference.

570k human-written English sentence pairs manually labeled for balanced classification with the labels entailment, contradiction, and neutral.

Paper ↗Download datasetSubmit a result ↵
§ 01 · Leaderboard

Best published scores.

8 results indexed across 1 metric. Shaded row marks current SOTA; ties broken by submission date.

Primary
accuracy · higher is better
accuracy· primary
8 rows
#ModelOrgSubmittedPaper / codeaccuracy
01GPT-4oAPIOpenAIMar 2023GPT-4 Technical Report92.60
02DeBERTa-v3-largeOSSMicrosoftNov 2021DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Tra…92.20
03Gemini UltraGoogle DeepMindDec 2023Gemini: A Family of Highly Capable Multimodal Models91.90
04Claude 3.5 SonnetAPIAnthropicJun 2024Claude 3.5 Sonnet Model Card91.80
05Llama 3.1 405BOSSMetaJul 2024The Llama 3 Herd of Models91.20
06Qwen2 72BAlibabaJul 2024Qwen2 Technical Report90.10
07Llama 3 70BOSSMetaJul 2024The Llama 3 Herd of Models89.70
08Mistral 7BMistral AIOct 2023Mistral 7B85.60
Fig 2 · Rows sorted by score within each metric. Shaded row marks SOTA. Dates reflect model or paper release where available, otherwise the date Codesota accessed the source.
§ 03 · Progress

2 steps
of state of the art.

Each row below marks a model that broke the previous record on accuracy. Intermediate submissions are kept in the leaderboard above; only SOTA-setting entries are re-listed here.

Higher scores win. Each subsequent entry improved upon the previous best.

SOTA line · accuracy
  1. Nov 18, 2021DeBERTa-v3-largeMicrosoft92.20
  2. Mar 15, 2023GPT-4oOpenAI92.60
Fig 3 · SOTA-setting models only. 2 entries span Nov 2021 → Mar 2023.
§ 04 · Literature

7 papers
tied to this benchmark.

Every paper below corresponds to at least one row in the leaderboard above. Click through for the arXiv preprint and, when available, the reference implementation.

§ 06 · Contribute

Have a score that beats
this table?

Submit a checkpoint and a reproduction script. We will run it, publish the score, and — if it takes the top — annotate the step on the progress chart with your name.

Submit a result ↵Read submission guide
What a submission needs
  • 01A public checkpoint or API endpoint
  • 02A reproduction script with frozen commit + seed
  • 03Declared evaluation environment (Python, deps)
  • 04One row per metric declared by this dataset
  • 05A contact so we can follow up on discrepancies