Skip to content

Wire Briefings Search

Fetching from the wire…

01

02

03

04

05

06

07

08

Wire Briefings Search

Entity trail

EvoHarness-RL

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

0

Findings

1

Edges

3

Sources

1

Corpus findings

2026-08-29 / news-researcherMeta's EvoHarness-RL Gets Qwen3-8B to 96.9% on ALFWorld, Within Two Points of Claude Opus 4.5arXiv 2608.05446 ('EvoHarness-RL: Learning Self-Evolving Runtime Harness for Long-Horizon LLM Agents') turns tool use from a hardcoded prompt into a learned runtime behavior, then applies cost-aware RL that teaches the agent when reading external state is worth the token budget. On ALFWorld, Qwen3-8B hits a 96.9% average success rate, beating SkillOS (80.2%) and SkillRL (89.9%); the same harness pushes Claude Opus 4.5 to 98.5%. For builders the claim is that harness design, not model size, is where the remaining points live.

Graph relationships

USES
Qwen3-8B
Qwen3-8B was tested with EvoHarness-RL runtime harness.
Source finding
BUILT BY
Meta
Meta developed EvoHarness-RL for learning runtime harness behavior in LLM agents.
Source finding
USES
Qwen3-8B -> EvoHarness-RL
Qwen3-8B was tested with EvoHarness-RL runtime harness.
Source finding

Source trail

Graph sources

entity graphfindings textkg edgeskg entities