paper-with-me

홈 › Papers

A$^2$TGPO: Agentic Turn-Group Policy Optimization with Adaptive Turn-level Clipping

2026-05-07 · Dingwei Chen, Zefang Zong, Zhipeng Ma, Leo Luo, Yang Li, Chengming Li, Peng Chen, Jie Jiang arxiv

Reinforcement learning for agentic large language models (LLMs) typically relies on a sparse, trajectory-level outcome reward, making it difficult to evaluate the contribution of individual tool-calls within multi-turn interactions. Existing approaches to such process credit assignment either depend on separate external process reward models that introduce additional consumption, or tree-based structural rollout that merely redistributes the outcome signal while constraining trajectory diversity. A promising alternative leverages the per-turn change in the policy's predicted probability of the ground-truth, termed Information Gain (IG), as an intrinsic process signal without an external evaluator. However, prior work on leveraging IG signals within the RL training loop faces three systematic challenges: normalizing across turns that face heterogeneous positional contexts can distort the relative standing of individual turns, accumulating a variable number of terms causes advantage magnitudes to drift with trajectory depth, and a fixed clipping range governs policy updates identically for turns with vastly different IG signals. In this paper, we propose A$^2$TGPO (Agentic Turn-Group Policy Optimization with Adaptive Turn-level Clipping), which retains IG as the intrinsic signal but re-designs how it is normalized, accumulated, and consumed: (i) turn-group normalization: normalizes IG within each (prompt, turn-index) group so that each turn is compared only against peers at the same interaction depth; (ii) variance-rescaled discounted accumulation: divides cumulative normalized IG by square root of accumulated terms to keep advantage magnitudes comparable across turn positions; and (iii) adaptive turn-level clipping: modulates each turn's clipping range based on its normalized IG, widening the update region for informative turns and narrowing it for uninformative ones.

📄 PDF Abstract BibTeX arXiv:2605.06200

Code (0)

등록된 구현이 없습니다.

Tasks

Reinforcement Learning

Similar Papers 제목 키워드 기반

Teacher-Guided Policy Optimization for On-Policy Reasoning Distillation under Large Policy Divergence

2026-05-13 · Xinyu Liu, Kechen Jiao, Chunyang Xiao, Runsong Zhao 외 arxiv

On-policy distillation (OPD) has become a promising paradigm for reasoning-oriented post-training of large language models (LLMs), especially when combined with reinforcement learning from verifiable rewards (RLVR). Exis…

Reinforcement Learning

TGPO: Temporal Grounded Policy Optimization for Signal Temporal Logic Tasks

2025-09-30 · Yue Meng, Fei Chen, Chuchu Fan arxiv

Learning control policies for complex, long-horizon tasks is a central challenge in robotics and autonomous systems. Signal Temporal Logic (STL) offers a powerful and expressive language for specifying such tasks, but it…

Reinforcement Learning

In-the-Flow Agentic System Optimization for Effective Planning and Tool Use

2025-10-07 · Zhuofeng Li, Haoxiang Zhang, Seungju Han, Sheng Liu 외 arxiv

Outcome-driven reinforcement learning has advanced reasoning in large language models (LLMs), but prevailing tool-augmented approaches train a single, monolithic policy that interleaves thoughts and tool calls under full…

Reinforcement Learning

PlanPO: Group Planning-Aware Policy Optimization for Multi-Turn Agentic LLMs

2026-08-18 · Dayang Liang, Liyuan He, Xuan Feng, Shuxin Li 외 arxiv

Group-relative policy optimization has emerged as a key paradigm for training agentic large language models (LLMs) on multi-turn interactive tasks. However, most existing variants fail to distinguish advantages among suc…

Incentivizing Temporal-Awareness in Egocentric Video Understanding Models

2026-03-28 · Zhiyang Xu, Tian Qin, Bowen Jin, Zhengfeng Lai 외 arxiv

Multimodal large language models (MLLMs) have recently shown strong performance in visual understanding, yet they often lack temporal awareness, particularly in egocentric settings where reasoning depends on the correct …

Reinforcement Learning