Model card
CQL (Conservative Q-Learning).
UC Berkeleyopen-sourceConservative Q-Learning — adds a regularizer to Q-values to penalize out-of-distribution actions
Kumar et al. NeurIPS 2020. One of the most widely cited offline RL baselines.
§ 02 · Benchmarks
Every benchmark CQL (Conservative Q-Learning) has a recorded score for.
| # | Benchmark | Area · Task | Metric | Value | Rank | Date | Source |
|---|---|---|---|---|---|---|---|
| 01 | d4rl-halfcheetah-medium | — | normalized_return | 44.0% | #3 | 2026-04-20 | source ↗ |
Rank column shows this model’s position vs all other models scored on the same benchmark + metric (competitors after the slash). #1 in red means current SOTA. Sorted by rank, then newest result.
§ 05 · Related models
Other UC Berkeley models scored on Codesota.
§ 06 · Sources & freshness
Where these numbers come from.
codesota-api
1
result
0 of 1 rows marked verified.