Differentiable Evolutionary Reinforcement Learning

• URL: http://arxiv.org/abs/2512.13399v1
• Date: Mon, 15 Dec 2025 14:50:08 GMT
• Title: Differentiable Evolutionary Reinforcement Learning
• Authors: Sitao Cheng, Tianle Li, Xuhan Huang, Xunjian Yin, Difan Zou,
• Abstract summary: We propose Differentiable Evolutionary Reinforcement Learning (DERL), a bilevel framework that enables the autonomous discovery of optimal reward signals.<n>DERL is differentiable in its metaoptimization: it treats the inner-loop validation performance as a signal to update the Meta-r via reinforcement learning.<n> Experimental results show that DERL achieves state-of-the-art performance on ALFWorld and ScienceWorld, significantly outperforming methods relying on rewards.
• Score: 41.96953381133274
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The design of effective reward functions presents a central and often arduous challenge in reinforcement learning (RL), particularly when developing autonomous agents for complex reasoning tasks. While automated reward optimization approaches exist, they typically rely on derivative-free evolutionary heuristics that treat the reward function as a black box, failing to capture the causal relationship between reward structure and task performance. To bridge this gap, we propose Differentiable Evolutionary Reinforcement Learning (DERL), a bilevel framework that enables the autonomous discovery of optimal reward signals. In DERL, a Meta-Optimizer evolves a reward function (i.e., Meta-Reward) by composing structured atomic primitives, guiding the training of an inner-loop policy. Crucially, unlike previous evolution, DERL is differentiable in its metaoptimization: it treats the inner-loop validation performance as a signal to update the Meta-Optimizer via reinforcement learning. This allows DERL to approximate the "meta-gradient" of task success, progressively learning to generate denser and more actionable feedback. We validate DERL across three distinct domains: robotic agent (ALFWorld), scientific simulation (ScienceWorld), and mathematical reasoning (GSM8k, MATH). Experimental results show that DERL achieves state-of-the-art performance on ALFWorld and ScienceWorld, significantly outperforming methods relying on heuristic rewards, especially in out-of-distribution scenarios. Analysis of the evolutionary trajectory demonstrates that DERL successfully captures the intrinsic structure of tasks, enabling selfimproving agent alignment without human intervention.

Related papers

• Test-driven Reinforcement Learning [1.1142354615369274]
  We propose a Test-driven Reinforcement Learning (TdRL) framework to tackle the reward design challenge in RL.<n>In TdRL, multiple test functions are used to represent the task objective rather than a single reward function.<n>We show that TdRL matches or outperforms handcrafted reward methods in policy training.
  arXiv  Detail & Related papers  (2025-11-11T06:58:52Z)
• Demystifying Reinforcement Learning in Agentic Reasoning [90.3737088727791]
  We conduct a comprehensive and systematic investigation to demystify reinforcement learning in agentic reasoning.<n>We highlight our key insights: (i) replacing stitched synthetic trajectories with real end-to-end tool-use trajectories yields a far stronger SFT.<n> Exploration-friendly techniques are crucial for agentic RL, such as clip higher, overlong reward shaping, and maintaining adequate policy entropy could improve the training efficiency.
  arXiv  Detail & Related papers  (2025-10-13T17:57:15Z)
• From Supervision to Exploration: What Does Protein Language Model Learn During Reinforcement Learning? [76.288870982181]
  Protein language models (PLMs) have advanced computational protein science through large-scale pretraining and scalable architectures.<n> reinforcement learning (RL) has broadened exploration and enabled precise multi-objective optimization in protein design.<n>We ask if RL improves sampling efficiency and, more importantly, if it reveals capabilities not captured by supervised learning.
  arXiv  Detail & Related papers  (2025-10-02T01:31:10Z)
• Agentic Reinforcement Learning with Implicit Step Rewards [92.26560379363492]
  Large language models (LLMs) are increasingly developed as autonomous agents using reinforcement learning (agentic RL)<n>We introduce implicit step rewards for agentic RL (iStar), a general credit-assignment strategy that integrates seamlessly with standard RL algorithms.<n>We evaluate our method on three challenging agent benchmarks, including WebShop and VisualSokoban, as well as open-ended social interactions with unverifiable rewards in SOTOPIA.
  arXiv  Detail & Related papers  (2025-09-23T16:15:42Z)
• Learning to Reason without External Rewards [100.27210579418562]
  Training large language models (LLMs) for complex reasoning via Reinforcement Learning with Verifiable Rewards (RLVR) is effective but limited by reliance on costly, domain-specific supervision.<n>We explore Reinforcement Learning from Internal Feedback (RLIF), a framework that enables LLMs to learn from intrinsic signals without external rewards or labeled data.<n>We propose Intuitor, an RLIF method that uses a model's own confidence, termed self-certainty, as its sole reward signal.
  arXiv  Detail & Related papers  (2025-05-26T07:01:06Z)
• RILe: Reinforced Imitation Learning [60.63173816209543]
  RILe (Reinforced Learning) is a framework that combines the strengths of imitation learning and inverse reinforcement learning to learn a dense reward function efficiently.<n>Our framework produces high-performing policies in high-dimensional tasks where direct imitation fails to replicate complex behaviors.
  arXiv  Detail & Related papers  (2024-06-12T17:56:31Z)
• REvolve: Reward Evolution with Large Language Models using Human Feedback [6.4550546442058225]
  Large language models (LLMs) have been used for reward generation from natural language task descriptions.<n>LLMs, guided by human feedback, can be used to formulate reward functions that reflect human implicit knowledge.<n>We introduce REvolve, a truly evolutionary framework that uses LLMs for reward design in reinforcement learning.
  arXiv  Detail & Related papers  (2024-06-03T13:23:27Z)
• Human-centric Reward Optimization for Reinforcement Learning-based Automated Driving using Large Language Models [15.11759379703718]
  One of the key challenges in current Reinforcement Learning (RL)-based Automated Driving (AD) agents is achieving flexible, precise, and human-like behavior cost-effectively.<n>This paper introduces an innovative approach that uses large language models (LLMs) to intuitively and effectively optimize RL reward functions in a human-centric way.
  arXiv  Detail & Related papers  (2024-05-07T09:04:52Z)
• REBEL: Reward Regularization-Based Approach for Robotic Reinforcement Learning from Human Feedback [61.54791065013767]
  A misalignment between the reward function and human preferences can lead to catastrophic outcomes in the real world.<n>Recent methods aim to mitigate misalignment by learning reward functions from human preferences.<n>We propose a novel concept of reward regularization within the robotic RLHF framework.
  arXiv  Detail & Related papers  (2023-12-22T04:56:37Z)
• Curricular Subgoals for Inverse Reinforcement Learning [21.038691420095525]
  Inverse Reinforcement Learning (IRL) aims to reconstruct the reward function from expert demonstrations to facilitate policy learning. Existing IRL methods mainly focus on learning global reward functions to minimize the trajectory difference between the imitator and the expert. We propose a novel Curricular Subgoal-based Inverse Reinforcement Learning framework, that explicitly disentangles one task with several local subgoals to guide agent imitation.
  arXiv  Detail & Related papers  (2023-06-14T04:06:41Z)
• Learning to Optimize for Reinforcement Learning [58.01132862590378]
  Reinforcement learning (RL) is essentially different from supervised learning, and in practice, these learneds do not work well even in simple RL tasks. Agent-gradient distribution is non-independent and identically distributed, leading to inefficient meta-training. We show that, although only trained in toy tasks, our learned can generalize unseen complex tasks in Brax.
  arXiv  Detail & Related papers  (2023-02-03T00:11:02Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.