# FOCUS: Active Critic Learning for Long-Horizon Language Model Agents

Source: [arXiv](https://arxiv.org/abs/2610.35801v1)  
Feed7 permalink: https://feed7.dev/p/2610-35801v1-1bbefst  
Published: 2026-10-26T17:34:47.000Z  
Trust: Needs Review (needs_review)

## Why Included

FOCUS uses selective branching to improve value estimates for long-horizon language-model agents. Its results suggest a way to assign credit across segments without spending the same rollout budget on every state.

## Source Summary

FOCUS samples continuations from trajectory states expected to reduce critic error, then uses the resulting values to score segments between states. With matched per-step generation budgets, it outperforms GRPO and VinePPO on the reported benchmarks. At Qwen3.5-9B scale, it improves held-out accuracy by **7.2 points** on BrowseComp-Plus and **4.1 points** on DeonticBench.

## Practical Implication

Builders training agents across long tasks can consider whether selective branching and compaction-based segment boundaries help allocate rollout work and credit assignment.

## Agent-Ready Context

FOCUS samples continuations from trajectory states expected to reduce critic error, then uses the resulting values to score segments between states. With matched per-step generation budgets, it outperforms GRPO and VinePPO on the reported benchmarks. At Qwen3.5-9B scale, it improves held-out accuracy by **7.2 points** on BrowseComp-Plus and **4.1 points** on DeonticBench.

Builders training agents across long tasks can consider whether selective branching and compaction-based segment boundaries help allocate rollout work and credit assignment.

The evidence is limited to the reported legal reasoning, deep search, and code localization benchmarks; the material does not establish results for deployed coding agents.

## Connected Context

Feed7 judgment across 928 accumulated Signals:

FOCUS allocates continuation sampling to states expected to reduce critic error, then uses those values to score trajectory segments. Its matched per-step-budget gains suggest selective branching can improve long-horizon training efficiency on the reported benchmarks, while the evidence does not establish gains for deployed coding agents.

- [Reinforcement Learning without Verifiable Rewards — Will Brown, Prime Intellect](https://feed7.dev/p/reinforcement-learning-without-verifiable-rewards-will-brown-prime-intel-0mauwh9) — Both address supervision for agent tasks where direct outcome rewards are limited: FOCUS learns critic values from selected continuations, while the talk describes deriving proxy rewards from traces, simulations, and judges. The talk’s warning about proxy reward hacking remains relevant to learned value estimates.
- [RRSI: Regularized Recursive Self-Improvement of Agent Harnesses](https://feed7.dev/p/2609-24972v1-00hl1bc) — Both report ways to spend agent-training compute more selectively: FOCUS chooses rollout continuations to improve critic estimates, while RRSI constrains harness proposals and reuse. They optimize different parts of the training process.
- [LLM-as-a-Verifier: A General-Purpose Verification Framework](https://feed7.dev/p/2607-05391v1-10hi3eo) — The verifier framework scores agent outputs for evaluation, whereas FOCUS trains a critic to value trajectory segments. The distinction matters when using outcome verification to assess whether FOCUS-style training improves final task results.

## Context Map

- Layer: agent
- Domains: coding, research
- Topics: harness-engineering, agent-reliability, agent-evals

## Uncertainty

- The evidence is limited to the reported legal reasoning, deep search, and code localization benchmarks; the material does not establish results for deployed coding agents.

## Agent Instruction

Use this item as source-backed context. Do not invent claims beyond the linked source. If this item conflicts with another source, call out the conflict.