MOSS: Self-Evolution through Source-Level Rewriting in Autonomous Agent Systems

• URL: http://arxiv.org/abs/2605.22794v2
• Date: Sat, 23 May 2026 13:08:16 GMT
• Title: MOSS: Self-Evolution through Source-Level Rewriting in Autonomous Agent Systems
• Abstract summary: MOSS is a system that performs self-rewriting at the source level on production agentic substrates.<n>It lifts a four-task mean grader score from 0.25 to 0.61 in a single cycle without human intervention.
• Score: 36.221246203666745
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Autonomous agentic systems are largely static after deployment: they do not learn from user interactions, and recurring failures persist until the next human-driven update ships a fix. Self-evolving agents have emerged in response, but all confine evolution to text-mutable artifacts -- skill files, prompt configurations, memory schemas, workflow graphs -- and leave the agent harness untouched. Since routing, hook ordering, state invariants, and dispatch live in code rather than in any text artifact, an entire class of structural failure is physically unreachable from the text layer. We argue that source-level adaptation is a fundamentally more general medium: it is Turing-complete, a strict superset of every text-mutable scope, takes effect deterministically rather than through base-model compliance, and does not erode under long-context drift. We present MOSS, a system that performs self-rewriting at the source level on production agentic substrates. Each evolution is anchored to an automatically curated batch of production-failure evidence and proceeds through a deterministic multi-stage pipeline; code modification is delegated to a pluggable external coding-agent CLI while MOSS retains stage ordering and verdicts. Candidates are verified by replaying the batch against the candidate image in ephemeral trial workers, then promoted via user-consent-gated, in-place container swap with health-probe-gated rollback. On OpenClaw, MOSS lifts a four-task mean grader score from 0.25 to 0.61 in a single cycle without human intervention.

Related papers

• Inference-Time Graph Engineering for Multi-Agent LLM Workflows [51.56038180639833]
  We introduce ReActNet, a training-free framework that compiles a query and a set of role-specialized agents into a sequence of directed communication graphs.<n>We show that ReActNet consistently improves over fixed-topology and learned-topology baselines while maintaining competitive inference cost.
  arXiv  Detail & Related papers  (2026-09-04T23:32:11Z)
• From Interaction Traces to Persistent Skills: Online Evolution for Computer-Use Agents [48.50635499112162]
  We present an online skill-evolution framework that converts interaction trajectories and evaluator feedback into a persistent, versioned library of reusable procedures.<n>We compare the full evolving-library system with a configuration-matched empty-library control across four OSWorld application domains.<n>In GIMP, provenance-aware analysis reveals retrieval across task-of-origin boundaries and revision, where repeated accepted edits fail to recover the originating task.
  arXiv  Detail & Related papers  (2026-09-04T08:29:45Z)
• 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)
• AEVAL: From Anecdotal to Deterministic Testing for Agentic Skill Workflows [1.6785141970297952]
  AEVAL (Agentic Evaluation) is a CI-integrated framework that replaces this practice with a deterministic, reproducible test pipeline for agentic skills.<n>Every skill change triggers a test event: the skill runs against a developer declared evaluation contract inside an automated executor.<n>A key ingredient is a structural separation between executor and grader, preventing a subtle but pervasive failure mode.
  arXiv  Detail & Related papers  (2026-07-16T21:33:05Z)
• TacoMAS: Test-Time Co-Evolution of Topology and Capability in LLM-based Multi-Agent Systems [55.81570336226014]
  We introduce TacoMAS, a test-time co-evolution framework for dynamic multi-agent systems.<n>TacoMAS formulates MAS inference as a task of online graph adaptation, where nodes represent agents with role-specific capabilities and edges define their communication topology.<n>Experiments on four benchmarks demonstrate that TacoMAS outperforms nearly 20 multi-agent baselines, achieving an average improvement of 13.3% over the strongest baseline.
  arXiv  Detail & Related papers  (2026-05-10T13:52:00Z)
• From Agent Loops to Deterministic Graphs: Execution Lineage for Reproducible AI-Native Work [0.10312968200748114]
  We introduce execution lineage: an execution model in which AI-native work is represented as a directed acyclic graph (DAG) of artifact-producing computations.<n>We compare execution-lineage replay against loop-centric update baselines on two controlled policy-memo update tasks.
  arXiv  Detail & Related papers  (2026-05-07T14:39:37Z)
• IMPACT-CYCLE: A Contract-Based Multi-Agent System for Claim-Level Supervisory Correction of Long-Video Semantic Memory [73.22944697933603]
  Existing pipelines produce opaque, end-to-end outputs that expose no intermediate state for inspection.<n>We present IMPACT-Cycle, a supervisory multi-agent system that reformulates long-video understanding as a multimodal iterative claim-level maintenance.
  arXiv  Detail & Related papers  (2026-04-22T03:03:33Z)
• Scaling Human-AI Coding Collaboration Requires a Governable Consensus Layer [22.42181408084751]
  Vibe coding produces correct, executable code at speed, but leaves no record of the structural commitments, dependencies, or evidence behind it.<n>We propose Agentic Consensus: a paradigm in which the consensus layer C, an operable world model, replaces code as the primary artifact of engineering.<n>We propose benchmark task families designed to measure whether consensus-based reduce human intervention compared to chat-driven baselines.
  arXiv  Detail & Related papers  (2026-04-20T06:53:32Z)
• Autogenesis: A Self-Evolving Agent Protocol [60.15939127351914]
  We introduce Autogenesis Protocol (AGP), a self evolution protocol that decouples what evolves from how evolution occurs.<n>We present Autogenesis System (AGS), a self-evolving multi-agent system that dynamically instantiates, retrieves, and refines protocol-registered resources during execution.
  arXiv  Detail & Related papers  (2026-04-16T14:04:06Z)
• Springdrift: An Auditable Persistent Runtime for LLM Agents with Case-Based Memory, Normative Safety, and Ambient Self-Perception [0.20305676256390928]
  We present Springdrift, a persistent runtime for long-lived LLM agents.<n>We introduce the term Artificial Retainer for this category.<n>This is a technical report on a systems design and deployment case study, not a benchmark-driven evaluation.
  arXiv  Detail & Related papers  (2026-04-06T13:14:37Z)
• OrgForge: A Multi-Agent Simulation Framework for Verifiable Synthetic Corporate Corpora [0.0]
  We present OrgForge, an open-source multi-agent simulation framework that enforces a strict physics-cognition boundary.<n>An actor-local clock enforces causal timestamp correctness across all artifact types.<n>Running a N-day simulation, OrgForge produces interleaved Slack threads, tickets, Confluence pages, Git pull requests, and emails, all traceable to a shared, immutable event log.
  arXiv  Detail & Related papers  (2026-03-16T09:02:24Z)
• SCOPE: Prompt Evolution for Enhancing Agent Effectiveness [53.75986399936395]
  Large Language Model (LLM) agents are increasingly deployed in environments that generate massive, dynamic contexts.<n>While agents have access to this context, their static prompts lack the mechanisms to manage it effectively.<n>We introduce textbfSCOPE (Self-evolving Context Optimization via Prompt Evolution)<n>We propose a Dual-Stream mechanism that balances tactical specificity (resolving immediate errors) with strategic generality (evolving long-term principles)
  arXiv  Detail & Related papers  (2025-12-17T12:25:05Z)
• SelfAug: Mitigating Catastrophic Forgetting in Retrieval-Augmented Generation via Distribution Self-Alignment [49.86376148975563]
  Large language models (LLMs) have revolutionized natural language processing through their capabilities in understanding and executing diverse tasks.<n> supervised fine-tuning, particularly in Retrieval-Augmented Generation (RAG) scenarios, often leads to catastrophic forgetting.<n>We propose SelfAug, a self-distribution alignment method that aligns input sequence logits to preserve the model's semantic distribution.
  arXiv  Detail & Related papers  (2025-09-04T06:50:47Z)

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.