Tool-calling evaluations for LLM agents are highly sensitive to implementation details such as random seeds and history handling, and two new techniques accelerate RL training with wall-clock speedup and no performance degradation.
Acebench: Who wins the match point in tool usage? arXiv preprint arXiv:2501.12851, 2025 a
7 Pith papers cite this work. Polarity classification is still indexing.
7
Pith papers citing it
citation-role summary
dataset 1
citation-polarity summary
roles
dataset 1polarities
use dataset 1representative citing papers
GEAR adaptively reweights GRPO advantages in LLM RL by using divergence spikes from self-distillation to define semantic segments and modulate local credit.
Entropy polarity is a signed token-level quantity derived from a first-order approximation of entropy change that predicts whether RL updates expand or contract policy entropy in LLM fine-tuning, revealing an asymmetry between high- and low-probability tokens.
LLMs show structural alignment bias by invoking semantically irrelevant tools when query attributes match tool parameters, revealed via SABEval dataset and mitigated by attention rebalancing.
MAVEN is a modular verification scaffold that lifts an open 120b model's tool-calling accuracy from 48% to 71% on MAVEN-Bench without retraining.
The paper delivers the first systematic review of self-evolving agents, structured around what components evolve, when adaptation occurs, and how it is implemented.
citing papers explorer
-
On Effectiveness and Efficiency of Agentic Tool-calling and RL Training
Tool-calling evaluations for LLM agents are highly sensitive to implementation details such as random seeds and history handling, and two new techniques accelerate RL training with wall-clock speedup and no performance degradation.
-
GEAR: Granularity-Adaptive Advantage Reweighting for LLM Agents via Self-Distillation
GEAR adaptively reweights GRPO advantages in LLM RL by using divergence spikes from self-distillation to define semantic segments and modulate local credit.
-
Entropy Polarity in Reinforcement Fine-Tuning: Direction, Asymmetry, and Control
Entropy polarity is a signed token-level quantity derived from a first-order approximation of entropy change that predicts whether RL updates expand or contract policy entropy in LLM fine-tuning, revealing an asymmetry between high- and low-probability tokens.
-
Do LLMs Know Tool Irrelevance? Demystifying Structural Alignment Bias in Tool Invocations
LLMs show structural alignment bias by invoking semantically irrelevant tools when query attributes match tool parameters, revealed via SABEval dataset and mitigated by attention rebalancing.
-
MAVEN: Improving Generalization in Agentic Tool Calling
MAVEN is a modular verification scaffold that lifts an open 120b model's tool-calling accuracy from 48% to 71% on MAVEN-Bench without retraining.
-
A Survey of Self-Evolving Agents: What, When, How, and Where to Evolve on the Path to Artificial Super Intelligence
The paper delivers the first systematic review of self-evolving agents, structured around what components evolve, when adaptation occurs, and how it is implemented.
- R2IF: Aligning Reasoning with Decisions via Composite Rewards for Interpretable LLM Function Calling