AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Agentic Monte Carlo: Simulating Reinforcement Learning for Black-Box Agents

Dae Yon Hwang, Raunaq Suri, Valentin Villecroze, Anthony L. Caterini, Jesse C. Cresswell, Noël Vouitsis, Brendan Leigh Ross

ICML 2026 · Reinforcement learning › RL from feedback · approach: Scaling RL with verifiable rewards · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use learned value-function resampling (AMC) instead of Best-of-15ReAct agent, Llama-3.2-11B · WebShopscore: 0.562 -> 0.625Improved
use AMC instead of Best-of-15ReAct agent, GPT-4.1-mini · WebShopscore: 0.403 -> 0.488Improved
use AMC instead of Best-of-15ReAct agent, Llama-3.2-11B · TextCraftscore: 0.296 -> 0.543Improved
use AMC instead of Best-of-15ReAct agent, GPT-5.1 · TextCraftscore: 0.889 -> 0.790Hurt
use AMC instead of Best-of-15ReflAct agent, Llama-3.1-8B · SciWorldscore: 0.347 -> 0.376Improved
add FoA heuristic-prompt value function SMC over Best-of-15Llama-3.2-11B · WebShopscore: 0.562 -> 0.580No clear effect
add a trained value function (vs prompted zero-shot value) (ablation)AMC, Llama-3.2-11B · WebShopscore: 0.556 -> 0.625Improved
add a trained value function (vs prompted zero-shot value) (ablation)AMC, GPT-5.1 · SciWorldscore: 0.506 -> 0.597Improved
add a trained value function (vs prompted zero-shot value) (ablation)AMC, GPT-5.1 · TextCraftscore: 0.815 -> 0.790Mixed
train value function on trajectories from the actual policy rather than a surrogate (ablation)AMC, GPT-4.1-mini · TextCraftscore: 0.790 -> 0.852Improved
train value function on trajectories from the actual policy rather than a surrogate (ablation)AMC, GPT-4.1-mini · WebShopscore: 0.467 -> 0.488Improved
pair cheap GPT-4.1-mini with AMC instead of GPT-5.1 Best-of-15cost comparison · SciWorldscore and cost: 0.533 -> 0.673; cost 0.18 -> 0.06Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy