ResearchGym - Continual Learning: leaderboard
ResearchGym task on rehearsal-free continual learning for foundation models on ImageNet-R/A, CIFAR-100 and CUB-200. A run's score is divided by the source paper's reported result (100 = matches it) and the best of three runs counts; higher is better.
Metric: Best-of-3 score normalised to the task's reference (SOTA) result, SOTA = 100. Source: anikethh.github.io. Saturation forecast: Around September 2027. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.5 | 98.5 | #152 |
| 2 | GPT-5.2 Codex (xHigh) | 97.9 | #185 (GPT-5.2 Codex) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=researchgym-continual-learning · How It Works · Data refreshed daily, snapshot 2026-11-12.