Reinforcement Learning for Long-Horizon Interactive LLM Agents
Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel pith:M3PBBCOOrecord.jsonopen to challenge →
read the original abstract
Interactive digital agents (IDAs) leverage APIs of stateful digital environments to perform tasks in response to user requests. While IDAs powered by instruction-tuned large language models (LLMs) can react to feedback from interface invocations in multi-step exchanges, they have not been trained in their respective digital environments. Prior methods accomplish less than half of tasks in sophisticated benchmarks such as AppWorld. We present a reinforcement learning (RL) approach that trains IDAs directly in their target environments. We formalize this training as a partially observable Markov decision process and derive LOOP, a data- and memory-efficient variant of proximal policy optimization. LOOP uses no value network and maintains exactly one copy of the underlying LLM in memory, making its implementation straightforward and as memory-efficient as fine-tuning a single LLM. A 32-billion-parameter agent trained with LOOP in the AppWorld environment outperforms the much larger OpenAI o1 agent by 9 percentage points (15% relative). To our knowledge, this is the first reported application of RL to IDAs that interact with a stateful, multi-domain, multi-app environment via direct API calls. Our analysis sheds light on the effectiveness of RL in this area, showing that the agent learns to consult the API documentation, avoid unwarranted assumptions, minimize confabulation, and recover from setbacks.
This paper has not been read by Pith yet.
Forward citations
Cited by 36 Pith papers
-
The Moltbook Files: A Harmless Slopocalypse or Humanity's Last Experiment
An AI-agent social platform generated mostly neutral content whose use in fine-tuning reduced model truthfulness comparably to human Reddit data, suggesting limited unique harm but flagging tail risks like secret leaks.
-
PRIME: Training Free Proactive Reasoning via Iterative Memory Evolution for User-Centric Agent
PRIME enables agents to proactively reason in user-centric tasks by iteratively evolving structured memories from interaction trajectories without gradient-based training.
-
ReCodeAgent: A Multi-Agent Workflow for Language-agnostic Translation and Validation of Large-scale Repositories
ReCodeAgent uses a multi-agent system to translate and validate large code repositories across multiple programming languages, achieving 60.8% higher test pass rates than prior neuro-symbolic and agentic methods on 11...
-
Orak: A Foundational Benchmark for Training and Evaluating LLM Agents on Diverse Video Games
Orak is a foundational benchmark providing training data, interfaces, and evaluation tools for LLM agents across diverse video game genres.
-
Group-in-Group Policy Optimization for LLM Agent Training
GiGPO adds a hierarchical grouping mechanism to group-based RL so that LLM agents receive both global trajectory and local step-level credit signals, yielding >12% gains on ALFWorld and >9% on WebShop over GRPO while ...
-
Rank-Then-Act: Reward-Free Control from Frame-Order Progress
RTA trains a VLM as a progress ordinal scorer via GRPO on shuffled expert frames and uses Spearman rank correlation with temporal indices as a bounded RL reward, matching or exceeding prior video reward methods on dis...
-
Diagnosing Task Insensitivity in Language Agents
The paper diagnoses task insensitivity in LLM agents as a cause of weak OOD generalization, links it to attention drift, and proposes Task-Perturbed NLL Optimization as a contrastive regularizer to improve task dependence.
-
Group-Graph Policy Optimization for Long-Horizon Agentic Reinforcement Learning
G2PO transforms linear trajectories into graphs, aggregates identical states for lower-variance value estimates, and uses edge-centric TD standardization, reporting up to 22.2% gains over GRPO on WebShop, ALFWorld, an...
-
When Denser Credit Is Not Enough: Evidence-Calibrated Policy Optimization for Long-Horizon LLM Agent Training
ECPO improves GiGPO by shrinking low-count action advantages and suppressing noisy anchor states, yielding +5.2/+7.3 success gains on ALFWorld/WebShop with Qwen2.5-1.5B models at negligible extra cost.
-
EvoTrainer: Co-Evolving LLM Policies and Training Harnesses for Autonomous Agentic Reinforcement Learning
EvoTrainer co-evolves LLM policies and training harnesses via empirical feedback to match or exceed human-engineered RL on math reasoning, code generation, and long-horizon software engineering.
-
Rewarding Beliefs, Not Actions: Consistency-Guided Credit Assignment for Long-Horizon Agents
ReBel uses belief-consistency supervision and belief-aware grouping to improve credit assignment in long-horizon RL for LLM agents, achieving up to 20.4 percentage points higher success and 2.1x better sample efficien...
-
What Really Improves Mathematical Reasoning: Structured Reasoning Signals Beyond Pure Code
Controlled experiments show structured reasoning traces and higher-density math-domain samples improve mathematical reasoning more than pure executable code, with internal routing patterns reflecting these data effects.
-
Resolving Action Bottleneck: Agentic Reinforcement Learning Informed by Token-Level Energy
ActFocus resolves the action bottleneck in agentic RL by reweighting token gradients toward action tokens using observed reward variance and an energy-based uncertainty term, outperforming PPO and GRPO by up to 65 per...
-
SOD: Step-wise On-policy Distillation for Small Language Model Agents
SOD reweights on-policy distillation strength step-by-step using divergence to stabilize tool use in small language model agents, yielding up to 20.86% gains and 26.13% on AIME 2025 for a 0.6B model.
-
Learning and Reusing Policy Decompositions for Hierarchical Generalized Planning with LLM Agents
HCL-GP learns parameterized policies and reuses extracted components to achieve 98% accuracy on AppWorld benchmark tasks for LLM agents, outperforming static synthesis by 15.8 points on challenges.
-
Skill1: Unified Evolution of Skill-Augmented Agents via Reinforcement Learning
Skill1 trains one policy to jointly evolve skill query generation, re-ranking, task solving, and distillation from a single task-success signal, with low-frequency trends crediting selection and high-frequency variati...
-
AEM: Adaptive Entropy Modulation for Multi-Turn Agentic Reinforcement Learning
AEM lifts entropy analysis to the response level and uses a derived uncertainty proxy to rescale advantages, enabling better exploration-exploitation balance and consistent gains over RL baselines on agent benchmarks.
-
FutureWorld: A Live Reinforcement Learning Environment for Predictive Agents with Real-World Outcome Rewards
FutureWorld enables RL training for predictive agents by recording rollouts at prediction time and replaying them with delayed real-world outcome rewards for policy updates.
-
FutureWorld: A Live Reinforcement Learning Environment for Predictive Agents with Real-World Outcome Rewards
FutureWorld is a live agentic RL environment that trains predictive agents using real-world outcome rewards and provides a daily benchmark for evaluating frontier agents.
-
A Survey on LLM-based Conversational User Simulation
A survey that introduces a taxonomy for LLM-based conversational user simulation, analyzes core techniques and evaluation methods, and identifies open challenges in the field.
-
Skill-SD: Skill-Conditioned Self-Distillation for Multi-turn LLM Agents
Skill-SD turns an agent's completed trajectories into dynamic natural-language skills that condition only the teacher in self-distillation, yielding 14-42% gains over RL and OPSD baselines on multi-turn agent benchmarks.
-
OASES: Outcome-Aligned Search-Evaluation Co-Training for Agentic Search
OASES co-trains search policies and evaluators to generate outcome-aligned process rewards, outperforming standard RL baselines on five multi-hop QA benchmarks.
-
World-Env: Leveraging World Model as a Virtual Environment for VLA Post-Training
World-Env replaces physical robot interactions with a world model-based virtual environment and VLM-guided rewards to enable efficient RL post-training for VLA models, showing gains with only five demonstrations per task.
-
WorldSample: Closed-loop Real-robot RL with World Modelling
WorldSample generates synthetic transitions from a post-trained world model grounded in real rollouts and uses Policy-Paced Learning to improve RL policies, reporting 28% higher success rates and 59% fewer training st...
-
SIRI: Self-Internalizing Reinforcement Learning with Intrinsic Skills for LLM Agent Training
SIRI trains LLM agents to discover, validate, and internalize reusable skills from their own rollouts without external generators or inference-time skill banks, yielding gains on ALFWorld and WebShop.
-
Skill1: Unified Evolution of Skill-Augmented Agents via Reinforcement Learning
Skill1 trains a single RL policy to co-evolve skill selection, utilization, and distillation in language model agents from one task-outcome reward, using low-frequency trends to credit selection and high-frequency var...
-
Skill1: Unified Evolution of Skill-Augmented Agents via Reinforcement Learning
Skill1 co-evolves skill selection, utilization, and distillation inside a single policy using only task-outcome reward, with low-frequency trends crediting selection and high-frequency variation crediting distillation...
-
On Training Large Language Models for Long-Horizon Tasks: An Empirical Study of Horizon Length
Longer action horizons bottleneck LLM agent training through instability, but training with reduced horizons stabilizes learning and enables better generalization to longer horizons.
-
FutureWorld: A Live Reinforcement Learning Environment for Predictive Agents with Real-World Outcome Rewards
FutureWorld closes the loop for predictive agents by storing predictions, backfilling real-world rewards when outcomes arrive, and replaying trajectories for policy updates, yielding consistent gains in accuracy and c...
-
FutureWorld: A Live Reinforcement Learning Environment for Predictive Agents with Real-World Outcome Rewards
FutureWorld is a modified verl-tool framework that enables delayed real-world outcome rewards for training LLM-based predictive agents, yielding consistent gains in accuracy, scoring, and calibration across three open...
-
SimpleSearch-VL: A Simple Recipe for Multimodal Agentic Deep Search
SimpleSearch-VL improves Qwen3-VL multimodal agent baselines by 15.8-16 points on average using 7K total training examples and reaches parity with Gemini-3-Pro on the 30B variant.
-
Shattering the Autoregressive Curse: Dynamic Epistemic Entropy Orchestrated Erasable Reinforcement Learning for LLMs
E³RL uses dynamic thresholds on epistemic entropy from autoregressive cross-entropy to enable erasable RL in LLM reasoning, reporting 5.349% and 6.514% gains on AIME for 4B and 8B models over prior SOTA.
-
Building Customer Support AI Agents at 100M-User Scale: An Evaluation-Driven Framework
An evaluation-driven framework for customer support AI agents at Nubank integrates context engineering, LLM judges, and A/B testing to deliver up to 37pp NPS gains and strong offline-online correlation across five pro...
-
GeoMathCode: Understanding Interleaved Math-Code Reasoning for Geometry Problem Solving
GeoMathCode interleaves math reasoning with programmatic code outputs for geometry problems in MLLMs and shows that reasoning steps and hierarchical code structures become disentangled in latent space after SFT.
-
AEM: Adaptive Entropy Modulation for Multi-Turn Agentic Reinforcement Learning
AEM adaptively modulates response-level entropy in agentic RL to improve credit assignment and exploration-exploitation balance, yielding gains on ALFWorld, WebShop, and SWE-bench.
-
A Survey of Scaling in Large Language Model Reasoning
A survey categorizing scaling in LLM reasoning across input size, steps, rounds, training, and future directions, noting that scaling can negatively affect performance.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.