AgentStack

arXiv:2604.02226 — When to ASK: Uncertainty-Gated Language Assistance for RL

by arXiv

3
researchVisit

Description

The paper addresses a real design question — **when to seek help** — but from the wrong direction for Forge. Our stack is LLM-native; MC Dropout cannot be applied to transformer inference.

Summary

Uncertainty-gated PPO+LM hybrid (Qwen2.5 0.5B–72B, N=100 MC Dropout passes, binary override not blending). Wrong paradigm for LLM-native agents; no in-domain improvement (baseline 0.93±0.26). U-shaped...

Tags

researchtypescriptopen-sourcepaperllm
License: CC BY 4.0Added: 2026-04-04

Related Tools