REVIEW 22 cited by
Stop Regressing: Training Value Functions via Classification for Scalable Deep RL
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Stop Regressing: Training Value Functions via Classification for Scalable Deep RL
read the original abstract
Value functions are a central component of deep reinforcement learning (RL). These functions, parameterized by neural networks, are trained using a mean squared error regression objective to match bootstrapped target values. However, scaling value-based RL methods that use regression to large networks, such as high-capacity Transformers, has proven challenging. This difficulty is in stark contrast to supervised learning: by leveraging a cross-entropy classification loss, supervised methods have scaled reliably to massive networks. Observing this discrepancy, in this paper, we investigate whether the scalability of deep RL can also be improved simply by using classification in place of regression for training value functions. We demonstrate that value functions trained with categorical cross-entropy significantly improves performance and scalability in a variety of domains. These include: single-task RL on Atari 2600 games with SoftMoEs, multi-task RL on Atari with large-scale ResNets, robotic manipulation with Q-transformers, playing Chess without search, and a language-agent Wordle task with high-capacity Transformers, achieving state-of-the-art results on these domains. Through careful analysis, we show that the benefits of categorical cross-entropy primarily stem from its ability to mitigate issues inherent to value-based RL, such as noisy targets and non-stationarity. Overall, we argue that a simple shift to training value functions with categorical cross-entropy can yield substantial improvements in the scalability of deep RL at little-to-no cost.
Forward citations
Cited by 22 Pith papers
-
WARP-RM: A Warp-Augmented Relative Progress Reward Model for Data Curation
WARP trains a reward model on time-warped successful demonstrations to produce frame-level progress estimates that upweight high-advantage chunks during behavior cloning, maintaining high success rates on suboptimal d...
-
Towards Efficient and Expressive Offline RL via Flow-Anchored Noise-conditioned Q-Learning
FAN achieves state-of-the-art offline RL performance on robotic tasks by anchoring flow policies and using single-sample noise-conditioned Q-learning, with proven convergence and reduced runtimes.
-
Chronos: Learning the Language of Time Series
Chronos pretrains transformer models on tokenized time series to deliver strong zero-shot forecasting across diverse domains.
-
WARP-RM: A Warp-Augmented Relative Progress Reward Model for Data Curation
Self-supervised relative progress velocities from time-warped demos let WARP-BC reweight action chunks so imitation policies stay robust as suboptimal teleop data is admitted.
-
WARP-RM: A Warp-Augmented Relative Progress Reward Model for Data Curation
Self-supervised relative progress from time-warped demos reweights BC action chunks, sustaining ~19/20 success and up to ~18× throughput on mixed-quality T-shirt folding where vanilla BC fails.
-
Superhuman AI for Generals.io Using Self-Play Reinforcement Learning
Self-play RL with a vision transformer policy, powered by a 10,000x faster JAX simulator, produces an agent that ranks #1 on the Generals.io leaderboard and wins 199-70 against top humans.
-
Survival Reinforcement Learning: Toward Scalable Self-Supervised RL
SRL is an online classification RL method extending survival value learning that matches CRL on manipulation and outperforms by 2-8x on long-horizon locomotion across robotic benchmarks.
-
Towards Efficient and Expressive Offline RL via Flow-Anchored Noise-conditioned Q-Learning
FAN simplifies expressive flow policies and distributional critics in offline RL via single-iteration behavior regularization and single-sample noise conditioning to claim SOTA performance with lower training and infe...
-
Hierarchical Behaviour Spaces
Hierarchical Behaviour Spaces uses linear combinations of reward functions to induce expressive behavior spaces in hierarchical RL, yielding strong performance on NetHack primarily through better exploration rather th...
-
What Does Flow Matching Bring To TD Learning?
Flow matching critics outperform monolithic ones in RL by 2x performance and 5x sample efficiency via test-time error recovery through integration and multi-point velocity supervision that preserves feature plasticity.
-
SERNF: Sample-Efficient Real-World Dexterous Policy Fine-Tuning via Action-Chunked Critics and Normalizing Flows
SERNF fine-tunes dexterous manipulation policies on real hardware by pairing normalizing-flow policies with action-chunked critics and conservative off-policy RL.
-
SERNF: Sample-Efficient Real-World Dexterous Policy Fine-Tuning via Action-Chunked Critics and Normalizing Flows
SERNF achieves sample-efficient real-world fine-tuning of multimodal dexterous policies by pairing exact-likelihood normalizing flow policies with action-chunked value critics.
-
Training Language Models to Self-Correct via Reinforcement Learning
SCoRe uses multi-turn online RL with regularization on self-generated traces to improve LLM self-correction, achieving 15.6% and 9.1% gains on MATH and HumanEval for Gemini models.
-
Relative Value Learning
A critic that learns antisymmetric value differences ∆(s_i,s_j)=V(s_i)−V(s_j) has a provably contracting Bellman operator and an unbiased advantage estimator, and PPO with this critic matches standard PPO on Atari.
-
World Value Models for Robotic Manipulation
World Value Model (WVM) integrates world models with value estimation to achieve SOTA Value-Order Correlation on expert and suboptimal robotic data and improves downstream policy performance.
-
Direct Advantage Estimation for Scalable and Sample-efficient Deep Reinforcement Learning
Extends DAE theory to POMDPs with minimal changes and introduces discrete latent dynamics to cut computational cost, with ALE experiments showing scalability and retained sample efficiency.
-
Distributional Value Estimation Without Target Networks for Robust Quality-Diversity
QDHUAC is a distributional, target-free QD-RL method that enables stable high-UTD training and competitive performance on Brax locomotion tasks using far fewer environment steps than prior approaches.
-
Low-Rank Adaptation for Critic Learning in Off-Policy Reinforcement Learning
LoRA applied to critics in SAC and FastTD3 reduces critic loss and yields best or competitive policy performance on most evaluated tasks.
-
RN-D: Discretized Categorical Actors for On-Policy Reinforcement Learning
A discretized categorical actor paired with a pre-LayerNorm residual MLP (RN-D) improves PPO's final performance and sample efficiency on continuous-control benchmarks.
-
D2 Actor Critic: Diffusion Actor Meets Distributional Critic
D2AC combines a diffusion actor with a distributional critic via fused distributional RL and clipped double Q-learning to reach state-of-the-art results on 18 hard control benchmarks including Humanoid, Dog, and Shadow Hand.
-
Naturalistic Computational Cognitive Science: Towards generalizable models and theories that capture the full range of natural behavior
Advocates integrating naturalistic paradigms and AI progress into cognitive science to develop generalizable models of natural behavior while retaining experimental control and theoretical insight.
-
Naturalistic Computational Cognitive Science: Towards generalizable models and theories that capture the full range of natural behavior
Position paper advocating integration of naturalistic paradigms and AI models to create generalizable theories of natural human behavior and cognition.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.