Google Open-Sources Harness2, a Recursive Agent Harnessing Framework

Google Cloud AI Research has open-sourced Harness². It’s a recursive agent harnessing framework that improves the code and recordsdata round a frozen mannequin. For builders, the pitch is easy. Your agent could get higher with out a higher mannequin.

TL;DR

Dimension: A framework, not a mannequin. 0 parameters educated; 8 editable harness elements. Model 0.1.0.
Runs on: Linux with Python 3.12 or 3.13 and uv. Mannequin runs name fashions via Vertex AI. GPU requirement: Not disclosed.
Efficiency: 1 refinement step improved all 4 mannequin and harness setups on all 3 benchmarks.
Greatest: GLM-5 below OpenCode rose from 64.2 to 75.7 on WorkBuddy-Bench (+11.5 factors).
Worst: GLM-5 below Codex gained just one.2 factors on JobBench (23.6 to 24.8).

What’s Harness²?

Harness² is a plug-and-play framework from Google Analysis that refines an agent’s harness throughout deployment. A harness is every thing across the mannequin: prompts, tools, memory, skills and orchestration. Harnessing is the method of optimizing that system. Mannequin weights by no means change.

The goal setting is the open world. Duties arrive unseen, and no verifier says whether or not a solution is correct. The research team argues that the majority self-improvement strategies assume labels or take a look at circumstances. Harness² is constructed to work with out them.

What are the 8 harness elements?

Harness² exposes 8 editable components, every mounted as a file:

System immediate
Guardrails
Reminiscence
Instrument steerage
Subagents
Abilities
Scripts
Plugins (middleware)

They cowl context building, the device interface, delegation and habits contained in the device loop. The format extends the 7-file break up from earlier Agentic Harness Engineering work by 1 element.

How does recursive agent harnessing work?

The framework runs two linked ranges of recursion:

Process-level recursion makes use of a contrastive Suggest, Probe, Compose step. First, the agent runs the duty as soon as with the bottom harness. An editor then proposes 2 candidates. The invariant candidate targets basic agent failures equivalent to planning or device use. The area candidate targets data and procedures for the present area. Each candidates run on the identical activity. The composer compares their trajectories towards the reference and retains solely edits backed by noticed habits.
Harnessing-level recursion occurs after every activity. A reflector shops the adopted harness plus an EXPERIENCE.md file of classes. This library guides future proposals. Every new activity nonetheless begins from the bottom harness.

The price range ok scales the search. Parallel mode probes ok invariant and ok area candidates, then makes ok compositions. Sequential mode chains ok refinement steps.

<div class="hd"><div class="dots"><i type="background:#4285F4"></i><i type="background:#EA4335"></i><i type="background:#FBBC04"></i><i type="background:#34A853"></i></div><h3>Harness<sup>2</sup> defined</h3></div>
<div class="sub">Google's recursive agent harnessing: the harness adjustments, the mannequin weights keep frozen.</div>

<div class="tabs">
<button class="tab on" data-t="p1">The way it works</button>
<button class="tab" data-t="p2">8 elements</button>
<button class="tab" data-t="p3">Benchmarks</button>
<button class="tab" data-t="p4">vs baselines</button>
</div>

<div class="pn on" id="h2-p1">
<div class="lvl">01 · Process-level recursion (inside 1 activity)</div>
<div class="movement">
<div class="nd" type="–c:#9aa0a6"><span class="tg" type="background:#9aa0a6">REFERENCE</span><b>Run base harness</b><small>H0 executes the duty as soon as</small></div>
<div class="nd" type="–c:#4285F4"><span class="tg" type="background:#4285F4">PROPOSE</span><b>2 candidates</b><small>Invariant + domain-specific edits</small></div>
<div class="nd" type="–c:#EA4335"><span class="tg" type="background:#EA4335">PROBE</span><b>Run each</b><small>Identical activity, similar mannequin, evaluate habits</small></div>
<div class="nd" type="–c:#FBBC04"><span class="tg" type="background:#FBBC04">COMPOSE</span><b>Merge what labored</b><small>Maintain edits backed by proof</small></div>
</div>
<div class="lvl">02 · Harnessing-level recursion (throughout duties)</div>
<div class="movement">
<div class="nd" type="–c:#4285F4"><span class="tg" type="background:#4285F4">EXECUTE</span><b>Ultimate harness</b><small>Deliverable returned</small></div>
<div class="nd" type="–c:#34A853"><span class="tg" type="background:#34A853">REFLECT</span><b>Evaluation the file</b><small>Which edits helped, and why</small></div>
<div class="nd" type="–c:#34A853"><span class="tg" type="background:#34A853">UPDATE</span><b>Enchancment library</b><small>Harness + EXPERIENCE.md</small></div>
<div class="nd" type="–c:#9aa0a6"><span class="tg" type="background:#9aa0a6">NEXT TASK</span><b>Restart at H0</b><small>Library guides new proposals</small></div>
</div>
<div class="be aware" id="h2-note">Press <b>Play</b> or <b>Step</b> to stroll via 1 Harness<sup>2</sup> enchancment step.</div>
<div class="ctl"><button class="btn pri" id="h2-play"> Play</button><button class="btn" id="h2-step">Step →</button><button class="btn" id="h2-reset">↺ Reset</button><div class="prog"><i id="h2-prog"></i></div></div>
<div class="fn">At ok=1: 1 reference + 2 probes + 1 last run per activity. Parallel mode probes ok invariant and ok area candidates; sequential mode chains ok steps.</div>
</div>

<div class="pn" id="h2-p2">
<div class="cg" id="h2-cg"></div>
<div class="be aware" id="h2-cnote" type="border-left-color:#4285F4">Click on a element to see its runtime function.</div>
<div class="lg"><span type="–c:#4285F4">Context building</span><span type="–c:#EA4335">Instrument interface</span><span type="–c:#FBBC04">Orchestration</span><span type="–c:#34A853">Middleware</span></div>
</div>

<div class="pn" id="h2-p3">
<div class="sel" id="h2-sel"></div>
<div class="bm" id="h2-bm"></div>
<div class="fn">Single enchancment step (ok=1). Means over 3 runs. LAB = Partial cross charge (%). Supply: Harness<sup>2</sup> paper, Tables 1 and a pair of.</div>
</div>

<div class="pn" id="h2-p4">
<div class="sel" id="h2-sel2"></div>
<div class="bx rk" id="h2-rk"></div>
<div class="fn">Gemini 3.5 Flash executor. Bar axis begins at 50. Meta-Harness is optimized offline on separate coaching duties. Scores as measured within the Harness<sup>2</sup> paper.</div>
</div>

<div class="ft"><span>Sources: arXiv 2610.14224 · github.com/google-research/harness2</span><a href="https://www.marktechpost.com" goal="_blank" rel="noopener">© Marktechpost</a></div>

(perform(){var f=doc.getElementById(‘mtp-h2x-frame’);window.addEventListener(‘message’,perform(e){if(e.supply===f.contentWindow&&e.information&&e.information.mtpH2xHeight){f.type.top=e.information.mtpH2xHeight+’px’;}});})();

How does Harness² carry out on benchmarks?

The analysis workforce examined 3 skilled benchmarks: WorkBuddy-Bench, JobBench and Harvey’s LAB. Executors have been Gemini 3.5 Flash and GLM-5 below OpenCode and Codex. The default editor was Claude Opus 4.8. Outcomes are means over 3 runs.

Setup (single step, ok=1)	WorkBuddy Avg	JobBench	LAB Partial
Gemini 3.5 Flash, OpenCode	72.4 → 77.5	31.5 → 36.0	79.9 → 81.9
Gemini 3.5 Flash, Codex	74.5 → 77.5	33.9 → 35.3	71.6 → 75.6
GLM-5, OpenCode	64.2 → 75.7	20.6 → 24.2	59.6 → 71.4
GLM-5, Codex	64.4 → 75.3	23.6 → 24.8	58.1 → 65.9

Supply: Harness² paper, Tables 1 and 2.

Weaker executors gained extra. After 1 step, GLM-5 beat base Gemini 3.5 Flash on WorkBuddy-Bench. Parallel recursion at ok=5 lifted Gemini below OpenCode to 80.8 on WorkBuddy. On WorkBuddy-Bench, 3 of 4 setups additionally used fewer steps per activity. Gemini below OpenCode fell from 25.4 to 21.9.

How does it evaluate with different harness strategies?

Characteristic	Harness²	Meta-Harness	ReasoningBank	Harness Scaling
Builders	Google Cloud AI Analysis, Northwestern	Stanford IRIS Lab and co-authors	Google Analysis and co-authors	Wang et al.
What adjustments	Full 8-component harness, per activity	Harness code, searched offline	Retrieved reminiscence; harness fastened	Harness, revised per occasion
Wants coaching break up	No	Sure	No	No
Learns throughout duties	Sure (library + classes)	Deploys 1 finest harness	Sure (reminiscence)	No
Code license	Apache-2.0	MIT	Not disclosed	Not disclosed
WorkBuddy Avg, Gemini/OpenCode*	77.5 (80.8 at ok=5)	76.4	73.4	75.3
JobBench, Gemini/OpenCode*	36.0	35.8	30.8	32.9
LAB Partial, Gemini/Codex*	75.6	76.5	74.8	67.9

*Scores as measured within the Harness² paper; the opposite strategies didn’t report these benchmarks themselves.

How do you put in and run Harness²?

The README targets Linux, Python 3.12 and uv. A neighborhood demo runs each recursion modes with out mannequin credentials:

Copy Code

uv venv --python 3.12 .venv
supply .venv/bin/activate
uv pip set up -e '.[render]'
harness2-demo --output runs/demo --k 2

Mannequin-backed runs want Node.js 20+, the Google Cloud CLI and a Vertex AI challenge. LAB wants Pandoc and Poppler. WorkBuddy-Bench wants Docker. Executors are OpenCode by default, or Codex with --substrate cx.

Key Takeaways

Harness² rewrites an agent’s harness at take a look at time; weights keep frozen.
1 step lifted GLM-5 by 11.5 factors on WorkBuddy-Bench.
Suggest, Probe, Compose replaces ground-truth labels with behavioral contrasts.
A classes library carries enhancements to future duties.
Count on additional runs per activity, and sandbox it your self.

Take a look at the Paper, GitHub Repo, Paper and Project Page. All credit score goes to the researcher of this challenge. Additionally, be happy to comply with us on Twitter and don’t overlook to affix our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well.

The put up Google Open-Sources Harness2, a Recursive Agent Harnessing Framework appeared first on MarkTechPost.

Google Open-Sources Harness2, a Recursive Agent Harnessing Framework

TL;DR

What’s Harness²?

What are the 8 harness elements?

How does recursive agent harnessing work?

How does Harness² carry out on benchmarks?

How does it evaluate with different harness strategies?

How do you put in and run Harness²?

Key Takeaways

No agenda, no rush, no leaving San Francisco: Benchmark’s companions on investing in a frenzied market

Converter

Editors Pick

Newsletter

Categories

Related Posts

Leave a Comment Cancel Reply

Latest