LAMBADA (Accuracy): leaderboard
LAMBADA (Paperno et al., 2016): predict the last word of a narrative passage, chosen so that people can guess it from the whole passage but not from the last sentence alone. Archived Papers With Code table of last-word accuracy as each paper reported it, with zero-, one- and few-shot rows; higher is better.
Metric: Accuracy. Source: huggingface.co. Saturation forecast: Estimated already saturated. 34 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | OPT-175B | 75.59 | #1507 |
| 2 | gpt-j-6B | 69.7 | #1589 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=lambada-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-05.