Auto Research with Specialist Agents Develops Effective and Non-Trivial Training Recipes

• URL: http://arxiv.org/abs/2605.05724v1
• Date: Thu, 07 May 2026 06:13:43 GMT
• Title: Auto Research with Specialist Agents Develops Effective and Non-Trivial Training Recipes
• Abstract summary: We study auto research as a closed empirical loop driven by external measurement.<n>Each submitted trial carries a hypothesis, an executable code edit, an evaluator-owned outcome, and feedback that shapes the next proposal.<n>We instantiate this loop with specialist agents that partition recipe surfaces and share measured lineage across trials.
• Score: 24.516406575400016
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: We study auto research as a closed empirical loop driven by external measurement. Each submitted trial carries a hypothesis, an executable code edit, an evaluator-owned outcome, and feedback that shapes the next proposal. The output is not a generated paper or a single model checkpoint, but an auditable trajectory of proposals, code diffs, experiments, scores, and failure labels. We instantiate this loop with specialist agents that partition recipe surfaces and share measured lineage across trials. The central empirical finding is that lineage feedback lets agents turn evaluator outcomes, including crashes, budget overruns, size failures, and accuracy-gate misses, into later program-level recipe edits rather than one-shot suggestions. Across 1,197 headline-run trials plus 600 Parameter Golf control trials after one-time setup and launch, humans did not choose proposals, edit recipes, override scores, or repair failed trials during the search. In the three headline runs, the same submitted-trial loop reduces Parameter Golf validation bpb by $0.81\%$, raises NanoChat-D12 CORE by $38.7\%$, and reduces CIFAR-10 Airbench96 wallclock by $4.59\%$, with each task measured by its own external evaluator and legality checks. The trace includes a strict architecture-domain audit of 157 headline-run submissions and program rewrites such as a NanoChat attention-kernel path change. Within this scope the loop autonomously writes code, submits experiments, absorbs feedback, applies and combines known techniques inside each environment, and improves public starting recipes.

Related papers

• Guardrailed Meta-Agent Loops: Stress-Testing Policy Pinning, Budget Bounds, and Crash Recovery [7.414694666903067]
  GuardrailLoop is a simulation-based testbed that makes three operational contracts jointly testable.<n>A hash-pinned policy fixes goals, scope, evaluation identity, budget, and release conditions.<n>It separates useful adaptation, state recovery, and repeated execution.
  arXiv  Detail & Related papers  (2026-09-10T21:22:08Z)
• From Experiments to Decisions: Reusing Evidence in Autonomous Coding Research [0.0]
  We reconstruct how evidence is reused in a 400-task NeuroGolf campaign.<n>A numerical counterexample exposes an overbroad exclusion.<n>Other episodes distinguish what failed, incomplete and revised programs justify doing next.
  arXiv  Detail & Related papers  (2026-09-10T03:48:31Z)
• Beneath the Diff: Diagnosing and Mitigating Algorithmic Mode Collapse in Code-Level Autonomous Research Loops [11.99667127177886]
  We show that repeated metric-driven code editing leads to genuine improvements that generalize beyond the loop.<n>We propose a lightweight mitigation that combines category-coverage reweighting, persistent edit memory, and a validation gate.
  arXiv  Detail & Related papers  (2026-08-31T08:11:19Z)
• Fantastic Adaptive Taxonomies and How to Use Them [100.20614173157708]
  AdaMAST converts execution traces into a compact, evidence-grounded failure taxonomy.<n>No code is hand-authored, no trace human-annotated.<n>In agent-system search, taxonomy-coded diagnoses of failed candidates outperform free-form.
  arXiv  Detail & Related papers  (2026-07-17T17:41:16Z)
• Interventional Causal Circuits for Safe Robot Action Testing and Failure Recovery [4.492262382019958]
  Formal testing of motion parameters is computationally expensive, and the cost scales poorly with the dimensionality of the action space.<n>We propose a closed-loop framework that couples a Joint Probability Tree (JPT) with a Causal Circuit derived from a Marginal-Deterministic Variable Tree.<n>The framework validates tractability of all interventional queries before the robot begins operating, and out-of-support candidates are detected and excluded from correction automatically.
  arXiv  Detail & Related papers  (2026-07-16T10:41:20Z)
• RouterVLA: Turning Smoke Tests into Supervision for Heterogeneous VLA Selection [5.188057876794867]
  We study whether pre-deployment evaluation rollouts can be reused to supervise policy selection.<n>We find that pre-deployment evaluation rollouts can be reused to supervise policy selection.
  arXiv  Detail & Related papers  (2026-06-25T17:56:33Z)
• Where Do Deep-Research Agents Go Wrong? Span-Level Error Localization in Agent Trajectories [51.22051230894794]
  Evaluation based on final answers shows whether an agent succeeds, but not which parts of the trajectory make the answer unreliable.<n>We collect 2,790 real trajectories from two agent frameworks, three backbone models, and three benchmarks, convert raw logs into semantic spans, and annotate harmful error spans through expert review.<n>We propose DRIFT, a claim-centric auditing framework that tracks agent claims, checks their support in trajectory evidence, and marks spans where unsupported or conflicting claims affect the answer path.
  arXiv  Detail & Related papers  (2026-06-01T10:50:26Z)
• What to Test Next: Interpretable Coverage Gap Discovery in Driving VLMs [52.50210189669399]
  Driving vision-language models (VLMs) must accurately understand scenes across diverse conditions defined by Operational Design Domains (ODDs)<n>We propose SliceScorer, a deterministic scoring rule for missing-slice recommendation.<n>We show that SliceNav surfaces high-risk coverage gaps more effectively than prior slice-discovery methods.
  arXiv  Detail & Related papers  (2026-06-01T03:18:01Z)
• From Patches to Trajectories: Privileged Process Supervision for Software-Engineering Agents [56.31499185764872]
  Supervised fine-tuning (SFT) on long teacher trajectories is the dominant way to instill investigation and reasoning in open software-engineering (SWE) agents.<n>We propose Patches-to-Trajectories (P2T), which uses $pstar$ as privileged information during curation and formulates trajectory construction as bi-objective optimization over per-step effectiveness and trajectory length.
  arXiv  Detail & Related papers  (2026-05-21T04:54:55Z)
• Reinforced Agent: Inference-Time Feedback for Tool-Calling Agents [6.158612515104146]
  We introduce Helpfulness-Harmfulness metrics to measure the tradeoff between agent feedback and reviewer feedback.<n>We evaluate our approach on BFCL and Tau2-Bench (multi-turn stateful scenarios), achieving +5.5% on irrelevance detection and +7.1% on multi-turn tasks.<n>Our metrics reveal that reviewer model choice is critical: the reasoning model o3-mini achieves a 3:1 benefit-to-risk ratio versus 2.1:1 for GPT-4o.
  arXiv  Detail & Related papers  (2026-04-29T22:09:47Z)
• The Last Human-Written Paper: Agent-Native Research Artifacts [106.47848184955576]
  We introduce the Agent-Native Research Artifact (ARA), a protocol that replaces the narrative paper with a machine-executable research package.<n>Three mechanisms support the ecosystem: a Live Research Manager that captures decisions and dead ends during ordinary development; an ARA Compiler that translates legacy PDFs and repos into ARAs; and an ARA-native review system that automates objective checks so human reviewers can focus on significance, novelty, and taste.
  arXiv  Detail & Related papers  (2026-04-27T16:23:09Z)
• Correction and Corruption: A Two-Rate View of Error Flow in LLM Protocols [51.56484100374058]
  We propose a paired-outcome measurement interface for auditing a single protocol step on exact-match tasks.<n>For each instance, the interface records a baseline correctness bit and a post-step correctness bit.<n>These rates predict accuracy changes and define a reusable empirical interface testable across seeds, mixtures, and pipelines.
  arXiv  Detail & Related papers  (2026-04-20T13:25:40Z)
• Runtime Execution Traces Guided Automated Program Repair with Multi-Agent Debate [8.424102114588559]
  Automated Program Repair (APR) struggles with complex logic errors and silent failures.<n>Current LLM-based APR methods are mostly static, relying on source code and basic test outputs.<n>We propose TraceRepair, a multi-agent framework that leverages runtime facts as shared constraints for patch validation.
  arXiv  Detail & Related papers  (2026-04-03T02:23:25Z)
• DeepReviewer 2.0: A Traceable Agentic System for Auditable Scientific Peer Review [22.737940168374546]
  DeepReviewer2.0 is a process-controlled agentic review system built around an output contract.<n>On 134 ICLR2025 submissions under three fixed protocols, an emphun-finetuned 196B model running DeepReviewer2.0 outperforms Gemini-3.1-Pro-preview.
  arXiv  Detail & Related papers  (2026-03-03T09:02:17Z)
• Solver-in-the-Loop: MDP-Based Benchmarks for Self-Correction and Behavioral Rationality in Operations Research [19.31559944205485]
  Operations Research practitioners routinely debug infeasible models through an iterative process.<n>We introduce two benchmarks that place the textbfsolver in the evaluation loop<n>We find that domain-specific RLVR training enables an 8B model to surpass frontier APIs.
  arXiv  Detail & Related papers  (2026-01-28T20:02:44Z)
• Sample-Efficient Expert Query Control in Active Imitation Learning via Conformal Prediction [2.344992278528697]
  We present Conformalized Rejection Sampling for Active Imitation Learning (CRSAIL)<n>CRSAIL scores state novelty by the distance to the $K$-th nearest expert state.<n>It reduces total expert queries by up to 96% vs. DAgger and up to 65% vs. prior AIL methods.
  arXiv  Detail & Related papers  (2025-11-29T11:58:21Z)
• A Simple Solution for Offline Imitation from Observations and Examples with Possibly Incomplete Trajectories [122.11358440078581]
  offline imitation is useful in real-world scenarios where arbitrary interactions are costly and expert actions are unavailable. We propose Trajectory-Aware Learning from Observations (TAILO) to solve MDPs where only task-specific expert states and task-agnostic non-expert state-action pairs are available.
  arXiv  Detail & Related papers  (2023-11-02T15:41:09Z)
• Self-Evaluation Guided Beam Search for Reasoning [61.523627290397556]
  We introduce a stepwise self-evaluation mechanism to guide and calibrate the reasoning process of Large Language Model (LLM) We propose a decoding algorithm integrating the self-evaluation guidance via beam search. Our approach surpasses the corresponding Codex-backboned baselines in few-shot accuracy by $6.34%$, $9.56%$, and $5.46%$ on the GSM8K, AQuA, and StrategyQA.
  arXiv  Detail & Related papers  (2023-05-01T02:37:59Z)
• Certified Error Control of Candidate Set Pruning for Two-Stage Relevance Ranking [57.42241521034744]
  We propose the concept of certified error control of candidate set pruning for relevance ranking. Our method successfully prunes the first-stage retrieved candidate sets to improve the second-stage reranking speed.
  arXiv  Detail & Related papers  (2022-05-19T16:00:13Z)

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.