AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

TQL: Scaling Q-Functions with Transformers by Preventing Attention Collapse

Perry Dong, Kuo-Han Hung, Alexander Swerdlow, Dorsa Sadigh, Chelsea Finn

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: Transformer expressivity and learnability · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add attention entropy loss with target entropy (ablation)Transformer Q-function (26M) · cube-double (5 tasks)success rateImproved
guide entropy toward a target instead of fixed max-entropy penalty (ablation)TQL (26M) · cube-double (5 tasks)success rateImproved
use per-layer and per-token temperatures instead of a single temperature (ablation)TQL (26M) · cube-double (5 tasks)success rateImproved
scale critic from 0.4M to 26M parametersprior offline RL methods (MLP, flow, transformer) · OGBenchsuccess rate: average decrease of 10.6%Hurt
scale plain transformer critic from 7M to 26MTransformer baseline without TQL · OGBenchsuccess rateHurt
TQL vs floq on OGBench overall averageTQL · OGBench 25 taskssuccess rate: 34 -> 40Improved
TQL vs puzzle-3x3 floqTQL · puzzle-3x3success rate: 35 -> 32Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy