HarnessNet: Feedback-driven Hierarchical Joint Harness Evolution for Multi-agent Systems

Zhaoyuan Qiu; Sachith Seneviratne; Yu Xia; Saman Halgamuge

cs.AIOct 10, 2026

HarnessNet: Feedback-driven Hierarchical Joint Harness Evolution for Multi-agent Systems

Authors: Zhaoyuan Qiu, Sachith Seneviratne, Yu Xia, Saman Halgamuge

Organizations: University of Melbourne, Melbourne, Australia

Abstract

AI agents can evolve by using environment feedback to revise their harnesses, the software managing model inputs, tool use, memory, and execution, while keeping foundation models frozen. Existing studies focus on single agents, leaving joint harness evolution in multi-agent systems (MAS) underexplored. A single global optimizer faces growing context and workload as the system scales, while access restrictions may prevent sharing agent internals or system-wide feedback. We propose HarnessNet, a framework for hierarchical harness evolution in MAS. Its basic unit, Agentron, comprises an evolvable forward harness, a foundation model, and a local optimizer that can observe and modify only that harness. Agentrons form a layered graph through which environment feedback propagates backward, with upstream optimizers receiving node-specific feedback from the downstream layer. This decomposes joint harness evolution into smaller, coordinated tasks with separate contexts, reducing individual optimizers' context and workload while keeping access local. Optimizers need not access all agent trajectories or raw environment feedback. Experiments on AgentBench with multiple foundation models demonstrate the feasibility of training HarnessNet. Compared with a single global optimizer, HarnessNet achieves comparable performance gains while respecting local privacy and requiring fewer tokens and less time per optimizer call. A migration study shows that the evolved harnesses remain jointly effective after replacing the foundation models, suggesting transferability across models.

Figures & tables

Figure 1: (Left) Agentron; (Right) Agentron Expansibility

Figure 2: The dual-phase of HarnessNet (Left; some links are omitted for clarity) and the propagation of feedback (Right).

Figure 3: HarnessNet 3-Stage Task-aware Initialization Worlflow, with a template provided as reference.

Benchmark	ALFWorld			WebShop
Split	Train	Validation	Test	Train	Validation	Test
Case Num	168	42	109	320	80	200

Table 1: Experiment Data Splits

Figure 4: Training Curves and Test Results on AgentBench of Minimal HarnessNet Topologies.

Benchmark Domain	Test	Score (↑)	Harness Gain	Training Gain
ALF-World	Model-only	0.5688	-	-
	FH0	0.2202	-61.29%	-
	FH_starred	0.7798	+37.10%	+254.13%
WebShop	Model-only	0.5670	-	-
	FH0	0.5559	-1.96%	-
	FH_starred	0.6950	+22.57%	+25.02%

Table 2: Agentron Training Effects

Benchmark Domain	Test	Score (↑)	Harness Gain	Training Gain
ALF-World	Model-only	0.5688	-	-
	FH0	0.6239	+9.69%	-
	FH_starred	0.8532	+50.00%	+36.75%
WebShop	Model-only	0.5670	-	-
	FH0	0.5523	-2.59%	-
	FH_starred	0.5892	+3.92%	+6.68%

Table 3: 2-Node HarnessNet Training Effects

Benchmark Domain	Test	Score (↑)	Harness Gain	Training Gain
ALF-World	Model-only	0.1193	-	-
	FH0	0.1651	+38.39%	-
	FH_starred	0.6972	+484.41%	+322.29%
WebShop	Model-only	0.4083	-	-
	FH0	0.0624	-84.72%	-
	FH_starred	0.5875	+43.89%	+841.51%

Table 4: 4-Node HarnessNet Training Effects

Panel of Figure 5 — Figure 5: Validation (Up) and Test Scores (Bottom) on ALFWorld across Different Foundation Models and Identical 4-node HarnessNet FH Snapshots Trained with Qwen3.5-9B.

Panel of Figure 6 — Figure 6: (Left) Illustration of Hierarchical Approach & Global Optimizer Approach; (Right) Training Performance.

Appendix figures & tables21 assets

Supplementary material from the paper’s appendix.

Appendix

Profile field	ALFWorld	WebShop
Objective	Complete a household manipulation goal through a sequence of environment actions.	Find and purchase a product satisfying the stated attributes and budget.
Observation	Raw text containing the task goal, environment feedback, and current admissible actions.	Raw text containing the shopping instruction or current page and the currently available actions.
Action interface	Exactly one required take_action call with a nonempty action string selected from the admissible actions.	Exactly one published search_action(keywords) or click_action(value) call valid for the current page.
Last-node output	One admissible action string; the adapter wraps it as take_action .	One object containing tool_name and arguments ; the adapter validates and submits the native tool-call envelope.
Step limit	15	20
Primary metric	Success rate, with success iff result.result == 1 .	Mean final reward, with per-case reward in $[0,1]$ and full success at result.result == 1.0 .

Appendix

Table A1 : Public benchmark profiles supplied to the HarnessNet initializer.

File	Initializer treatment	Purpose
harness.json	Controller-owned; scoped metadata editable	Declares node identity, topology neighbors, entry point, contract versions, template revision, and path/field allowlists. The materializer writes the identity fields; the builder may change only the allowed role metadata.
pyproject.toml	Fixed	Defines the self-contained Python package, supported interpreter version, dependencies, and test discovery settings.
README.md	Fixed	Records the neutral Contract 1.2 baseline, invocation command, and editing boundary for the repository.
ROLE.md	Builder-owned	Provides the auditable, node-specific role description derived from the accepted initialization plan.
prompts/system.md	Builder-owned	Supplies the node’s runtime system instruction and its plan-derived inductive bias.
prompts/forward.md	Fixed	Retains a future trainable prompt extension point; the neutral FH0 strategy does not inject this file into the model request.

Appendix

Table A2 : Files in the canonical harness template and their initializer-time roles.

Parameter	Qwen3.8-27b	Qwen3.5-9b	Gemma4-8b-e4b	Ornith-1.5-9B
Resource	Hugging Face	Hugging Face	Hugging Face	Hugging Face
Quantization	Q4	Q4	Q8	Q4
Backend	Ollama	Ollama	Ollama	Ollama
num_ctx	32,768	20,480	20,480	20,480
num_predict	4,096	2,048	2,048	2,048
temperature	0	0	0	0

Appendix

Table B1 : Ollama model and execution controls used in the selected HarnessNet runs. The two Qwen3.8-27b limits are the only generation-control differences among the four configurations.

Field	Value
Forward model	Qwen3.8-27b
Benchmark domain	ALFWorld
Topology	Agentron $[1]$
Train dataset	alfworld-env_train ( $n=168$ )
Validation dataset	alfworld-env_train ( $n=42$ )
Test dataset	alfworld-std , new_std split ( $n=109$ )

Appendix

Table C1 : Key metadata for ALFWorld-Agentron training.

Checkpoint	Train Score	Validation Score	Test Score
Model-only			56.88%
FH0	59.18%	40.48%	22.02%
FH1	39.39%	52.38%
FH2	55.56%	66.67%
FH3	58.54%	66.67%
FH4	76.47%	83.33%

Appendix

Table C2 : ALFWorld-Agentron train, validation, and held-out test records.

Field	Value
Forward model	Qwen3.8-27b
Benchmark domain	WebShop
Topology	Agentron $[1]$
Train dataset	webshop-env_train ( $n=320$ )
Validation dataset	webshop-env_train ( $n=80$ )
Test dataset	webshop-std , standard split ( $n=200$ )

Appendix

Table C3 : Key metadata for WebShop-Agentron training.

Checkpoint	Train Score	Validation Score	Test Score
Model-only			0.5670
FH0	0.4789	0.5357	0.5559
FH1	0.5047	0.6384
FH2	0.6691	0.6615
FH3	0.7078	0.7223
FH4	0.6863	0.7010

Appendix

Table C4 : WebShop-Agentron train, validation, and held-out test records.

Field	Value
Forward model	Qwen3.8-27b
Benchmark domain	ALFWorld
Topology	2Node HarnessNet $[1,1]$
Train dataset	alfworld-env_train ( $n=168$ )
Validation dataset	alfworld-env_train ( $n=42$ )
Test dataset	alfworld-std , new_std split ( $n=109$ )

Appendix

Table C5 : Key metadata for ALFWorld-2Node HarnessNet training.

Checkpoint	Train Score	Validation Score	Test Score
Model-only			56.88%
FH0	44.44%	57.14%	62.39%
FH1	60.42%	71.43%
FH2	45.21%	47.62%
FH3	89.47%	85.71%
FH4	91.67%	85.71%

Appendix

Table C6 : ALFWorld-2Node HarnessNet train, validation, and held-out test records.

Field	Value
Forward model	Qwen3.8-27b
Benchmark domain	WebShop
Topology	2Node HarnessNet $[1,1]$
Train dataset	webshop-env_train ( $n=320$ )
Validation dataset	webshop-env_train ( $n=80$ )
Test dataset	webshop-std , standard split ( $n=200$ )

Appendix

Table C7 : Key metadata for WebShop-2Node HarnessNet training.

Checkpoint	Train Score	Validation Score	Test Score
Model-only			0.5670
FH0	0.4807	0.5852	0.5523
FH1	0.6021	0.5998
FH2	0.5814	0.6450
FH3	0.5429	0.5885
FH4	0.6282	0.6054

Appendix

Table C8 : WebShop-2Node HarnessNet train, validation, and held-out test records.

Field	Value
Forward model	Qwen3.5-9b
Benchmark domain	ALFWorld
Topology	4Node HarnessNet $[1,2,1]$
Train dataset	alfworld-env_train ( $n=168$ )
Validation dataset	alfworld-env_train ( $n=42$ )
Test dataset	alfworld-std , new_std split ( $n=109$ )

Appendix

Table C9 : Key metadata for ALFWorld-4Node HarnessNet training.

Checkpoint	Train Score	Validation Score	Test Score
Model-only			11.93%
FH0	9.09%	19.05%	16.51%
FH1	34.43%	33.33%
FH2	33.33%	28.57%
FH3	39.39%	38.10%
FH4	42.86%	38.10%

Appendix

Table C10 : ALFWorld-4Node HarnessNet train, validation, and held-out test records.

Field	Value
Forward model	Gemma4-8b-e4b
Benchmark domain	WebShop
Topology	4Node HarnessNet $[1,2,1]$
Train dataset	webshop-env_train ( $n=320$ )
Validation dataset	webshop-env_train ( $n=80$ )
Test dataset	webshop-std , standard split ( $n=200$ )

Appendix

Table C11 : Key metadata for WebShop-4Node HarnessNet training.

Checkpoint	Train Score	Validation Score	Test Score
Model-only			0.4083
FH0	0.1210	0.0963	0.0624
FH1	0.4755	0.5298
FH2	0.5984	0.6167	0.5875
FH3	0.4837	0.5379
FH4	0.5710	0.4727

Appendix

Table C12 : WebShop-4Node HarnessNet train, validation, and held-out test records.

Field	Value
Forward models	Qwen3.5-9B; Ornith1.5-9B; gemma4-8B-E4B
Benchmark domain	ALFWorld
Topology	4Node HarnessNet $[1,2,1]$
Validation dataset	alfworld-env_train ( $n=42$ )
Test dataset	alfworld-std , new_std split ( $n=109$ )

Appendix

Table D1 : Key metadata for the migration study.

Checkpoint/Model	Qwen3.5-9B	Ornith1.5-9B	gemma4-8B-E4B
FH0	19.05%	16.67%	16.67%
FH1	33.33%	19.05%	16.67%
FH2	28.57%	23.81%	7.14%
FH3	38.10%	28.57%	30.95%
FH4	38.10%	33.33%	38.10%
FH5	76.19%	57.14%	78.57%

Appendix

Table D2 : Migration-study validation and held-out test pass rates.

	Hierarchical Optimizer			Global Optimizer
Checkpoint	Train Score	Validation Score	Test Score	Train Score	Validation Score	Test Score
Model-only			11.93%			11.93%
FH0	9.09%	19.05%	16.51%	9.09%	19.05%	16.51%
FH1	34.43%	33.33%		32.20%	28.57%
FH2	33.33%	28.57%		23.08%	30.95%
FH3	39.39%	38.10%		32.20%	45.24%

Appendix

Table E1 : ALFWorld train, validation, and held-out test pass rates under the hierarchical and global optimizers.

	Hierarchical Optimizer			Global Optimizer
Checkpoint	Train Score	Validation Score	Test Score	Train Score	Validation Score	Test Score
Model-only			0.4083			0.4083
FH0	0.1210	0.0963	0.0624	0.1210	0.0963	0.0624
FH1	0.4755	0.5298		0.3883	0.3802
FH2	0.5984	0.6167	0.5875	0.4811	0.4730
FH3	0.4837	0.5379		0.4837	0.5554

Appendix

Table E2 : WebShop train, validation, and held-out test AvgScore under the hierarchical and global optimizers.

	Hierarchical Optimizer				Global Optimizer
Backward	Node	Input Token	Output Token	Optimizer Time (s)	Input Token	Output Token	Optimizer Time (s)
1	l2-last	4,557,381	31,320	1,427	10,467,460	40,902	1,719
	l1-hidden_a	4,440,628	24,540	1,304
	l1-hidden_b	6,460,733	25,899	1,366
	l0-first	7,133,103	25,820	1,467
2	l2-last	7,027,571	46,165	1,912	13,542,625	42,660	2,056

Appendix

Table E3 : ALFWorld optimizer cost by backward round.

Backward	Node	Input Token	Output Token	Optimizer Time (s)	Input Token	Output Token	Optimizer Time (s)
	Hierarchical Optimizer				Global Optimizer
1	l2-last	3,971,824	18,694	707	20,197,746	44,424	2,378
	l1-hidden_a	4,240,691	18,872	776
	l1-hidden_b	2,977,722	14,114	603
	l0-first	5,586,749	22,601	870
2	l2-last	2,659,254	15,242	627	9,745,839	25,382	1,373

Appendix

Table E4 : WebShop optimizer cost by backward round.

Agent Harness Optimization

Agent harness optimization improves the prompts, tools, memory, workflows, and control code that shape how AI agents act.

See all 185 papers

HarnessX: A Composable, Adaptive, and Evolvable Agent Harness Foundry
Jun 12, 2026Tingyang Chen, Shuo Lu, Kang Zhao +11LLM Agent Orchestration Agent Harness Optimization
Darwin Agent Team
Harmony Harness: Multi-Objective Self-Evolution of Agent Harnesses
Oct 10, 2026Weichen Yu, Lesly Miculicich, Bhavana Dalvi Mishra +5Agent Harness Optimization LLM Agent Harnesses
Google Cloud AI Research · Carnegie Mellon University · Google DeepMind
HarnessCompass: Guiding Automatic Harness Evolution toward Generalizable and Effective Agent Harnesses
Aug 3, 2026Luan Zhang, Ruochen Zhou, Dandan Song +9Agent Harness Optimization LLM Agents
Beijing Institute of Technology, China · City University of Hong Kong, China · Independent, China

Jun 12, 2026J/K move · Enter open · S save

HarnessX: A Composable, Adaptive, and Evolvable Agent Harness Foundry

Tingyang Chen, Shuo Lu, Kang Zhao, Weicheng Meng+10

Darwin Agent Team

LLM Agent Orchestration Agent Harness Optimization LLM Agent Workflow Optimization LLM Agent Self-Improvement LLM Agent Harnesses Self-Evolving Agents

Open paper AskPDF arXiv

AI agent performance depends critically on the runtime harness, comprising the prompts, tools, memory, and control flow that mediate how a model observes, reasons, and acts. Yet today's harnesses remain largely hand-crafted and static: each new model or task still demands bespoke scaffolding, and the rich traces produced during execution are rarely distilled back into systematic improvement. We introduce HarnessX, a foundry for composable, adaptive, and evolvable agent harnesses. HarnessX assembles typed harness primitives via a substitution algebra, adapts them through AEGIS, a trace-driven multi-agent evolution engine grounded in an operational mirror between symbolic adaptation and reinforcement learning, and closes the harness-model loop by turning trajectories into both harness updates and model training signal. Across five benchmarks (ALFWorld, GAIA, WebShop, tau^3-Bench, and SWE-bench Verified), HarnessX yields an average gain of +14.5% (up to +44.0%), with gains largest where baselines are lowest. These results suggest that agent progress need not come from model scaling alone: composing and evolving runtime interfaces from execution feedback is an actionable and complementary lever. The complete codebase will be open-sourced in a future release.

Abstract

Agent Harness Optimization

Explore similar work

HarnessX: A Composable, Adaptive, and Evolvable Agent Harness Foundry

Harmony Harness: Multi-Objective Self-Evolution of Agent Harnesses

HarnessCompass: Guiding Automatic Harness Evolution toward Generalizable and Effective Agent Harnesses