AI News/Wire summary
Harness^2 uses recursive self-improvement to evolve LLM agent harnesses
Key points
- Harness^2 is a plug-and-play framework with task-level and harnessing-level recursion for open-world agent tasks.
- Task-level step uses a contrastive Propose-Probe-Compose process that works without ground-truth verifiers, per the abstract.
- Reported gains reach 11.5 points on WorkBuddy-Bench, 4.5 on JobBench and 11.8 percentage points on LAB, tested on GLM-5 and Gemini 3.5 Flash.
Why it matters for marketers: Teams building agent workflows may get better results by tuning the execution harness around the model, though the gains are reported from one preprint.
Summary by the Z Media news desk, from the outlets' reporting. Read the full stories at the source.
Read the original
arXiv cs.AI: Harness^2: Recursive Agent Harnessing for an Open World
Related on Z Media
- SpaceXSI folds X Premium, Grok and Cursor into one five-tier subscriptionOur brief, via TechCrunch, 2 days ago
- Nvidia opens up Super VL, a model for always-on enterprise agentsOur brief, via The Decoder, 8 hr ago
- Reported AI agents misused exposed credentials to probe Amapvia The Information, yesterday
- Square launches a voice AI kiosk for restaurant ordering at $1,999Our brief, via The Verge, 2 days ago
- Shopify lets shoppers connect any AI agent to buy from its catalogOur brief, via The Verge, 2 days ago
- LFM2.5-24B-A2B: open-weight reasoning agent model for on-device usevia Hugging Face, 2 days ago