WHALE: A Simple Recipe for Joint Harness-Weight Optimization

• URL: http://arxiv.org/abs/2609.00196v1
• Date: Mon, 31 Aug 2026 18:12:55 GMT
• Title: WHALE: A Simple Recipe for Joint Harness-Weight Optimization
• Abstract summary: Weight-Harness Alternating LEarning (WHALE) is a recipe that alternates two phases: updating the model under the current harness, then searching for a better harness under the updated model.<n>Using Qwen3.5-2B/4B agents across three domains (search question answering, mathematical reasoning, and chess puzzles), WHALE outperforms weight-only, harness-only, and Fast-Slow Training by 4.15-24.38 percentage points in best mean@8 accuracy.
• Score: 48.550149284101536
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Agent performance depends jointly on the model parameters and the executable harness code that manages context and control flow. Optimizing either component in isolation can leave the system bottlenecked by its frozen counterpart: weight updates can change which harness is effective, while harness updates can change which model capabilities are exposed. Existing joint-adaptation methods optimize weights and textual prompts but leave the broader harness fixed. We propose Weight-Harness Alternating LEarning (WHALE), a simple recipe that alternates two phases: updating the model under the current harness, then searching for a better harness under the updated model. We instantiate these two phases with online rejection-sampling fine-tuning and Meta-Harness, respectively. When to switch is a key design choice: to separate real improvements from noise without over-optimizing against a changing counterpart, WHALE uses either fixed phase durations or an adaptive patience rule over training signals. Using Qwen3.5-2B/4B agents across three domains (search question answering, mathematical reasoning, and chess puzzles), WHALE outperforms weight-only, harness-only, and Fast-Slow Training by 4.15-24.38 percentage points in best mean@8 accuracy. Either component can be the bottleneck: harness search matches peak weight-only accuracy with far fewer rollouts in SearchQA, but improves math accuracy only after a weight update. Small interleaved updates also outperform stagewise weight-then-harness optimization in accuracy and rollout cost. The code is available at https://github.com/krafton-ai/WHALE.

Related papers

• HarnessBandit: Joint Learnability-Transferability Scheduling for Multi-Harness Agentic Reinforcement Learning [40.61476978001808]
  We develop HarnessBandit, an online scheduler that selects one harness per step.<n>We train Qwen3.5-2B across six harnesses on ClawGym and evaluate on PinchBench and ClawEval.
  arXiv  Detail & Related papers  (2026-09-12T06:22:45Z)
• Low-Rank Ternary Adaptation for Fine-Tuning Transformers [54.3230438745856]
  Ternary transformers offer extreme memory and compute efficiency.<n>Existing low-bit LoRA-based methods cannot directly fine-tune ternary weights.<n>We propose ternary multiplicative adaptation, which represents discrete updates of ternary weights.
  arXiv  Detail & Related papers  (2026-08-25T12:15:56Z)
• AutoSaddler: Automatic Harness Optimization with Durable Updates from Agent Execution Traces [61.51487329124135]
  AutoSaddler is a framework that formulates harness improvement as an offline learning problem.<n>It iteratively updates the harness using failure signals from mini-batches.<n>Experiments on GAIA2, SWE-Bench Pro, and Terminal-Bench 2.0 show that AutoSaddler substantially improves agent performance.
  arXiv  Detail & Related papers  (2026-08-24T09:45:08Z)
• Improving Neural Network Training by Decoupling the Magnitude and Direction of Weight Vectors [110.10775872372047]
  We propose emphMagnitude-Direction-Direction (MD) Decoupling.<n>It factorizes each weight into a fixed-norm direction on a hypersphere and learnable per-row and per-column magnitude gains, updated at separate learning rates.<n>Across both Adam and Muon, MD Decoupling improves on well-tuned baselines, transfers the optimal LR across model width without retuning, and continues to help at scale on large Mixture-of-Experts models.
  arXiv  Detail & Related papers  (2026-06-24T15:40:26Z)
• Meta-Harness: End-to-End Optimization of Model Harnesses [52.31507076660471]
  We introduce Meta-Harness, an outer-loop system that searches over harness code for large language model (LLM) applications.<n>On online text classification, Meta-Harness improves over a state-of-the-art context management system by 7.7 points while using 4x fewer context tokens.<n>On retrieval-augmented math reasoning, a single discovered harness improves accuracy on 200 IMO-level problems by 4.7 points on average across five held-out models.
  arXiv  Detail & Related papers  (2026-03-30T05:33:50Z)
• STABLE: Gated Continual Learning for Large Language Models [0.0]
  STABLE is a gated continual self editing framework that constrains forgetting during sequential updates.<n>Each candidate edit is evaluated against a stability budget using one of three metrics.<n>Experiments on the Qwen-2.5-7B model show that gating effectively mitigates forgetting while preserving adaptability.
  arXiv  Detail & Related papers  (2025-10-17T16:14:05Z)
• ABBA-Adapters: Efficient and Expressive Fine-Tuning of Foundation Models [10.17362679822278]
  Large Language Models have demonstrated strong performance across a wide range of tasks, but adapting them efficiently to new domains remains a key challenge.<n>We introduce ABBA, a new PEFT architecture that re parameterizes the update as a Hadamard product of two independently learnable low-rank matrices.<n>In contrast to prior work, ABBA fully decouples the update from the pre-trained weights, enabling both components to be optimized freely.
  arXiv  Detail & Related papers  (2025-05-20T11:43:25Z)
• NEAT: Nonlinear Parameter-efficient Adaptation of Pre-trained Models [26.808251361020066]
  Fine-tuning pre-trained models often yields state-of-the-art performance but is computationally expensive when updating all parameters.<n>We propose NEAT, a nonlinear PEFT approach that employs a lightweight neural network to learn a nonlinear transformation of the pre-trained weights.<n>Our theoretical analysis shows that NEAT achieves greater efficiency than LoRA while maintaining equivalent expressivity.
  arXiv  Detail & Related papers  (2024-10-02T17:29:23Z)
• AdamP: Slowing Down the Slowdown for Momentum Optimizers on Scale-invariant Weights [53.8489656709356]
  Normalization techniques are a boon for modern deep learning. It is often overlooked, however, that the additional introduction of momentum results in a rapid reduction in effective step sizes for scale-invariant weights. In this paper, we verify that the widely-adopted combination of the two ingredients lead to the premature decay of effective step sizes and sub-optimal model performances.
  arXiv  Detail & Related papers  (2020-06-15T08:35:15Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.