state snapshotting

This work addresses the prevalent issue in autonomous coding agents that prematurely declare lifecycle states—such as “DONE”—without verification during multi-step software tasks, often leading to erroneous progression. To mitigate this, the authors propose Proof-or-Stop, a model-agnostic and platform-neutral trusted control layer that strictly gates state transitions only when fresh, traceable, and mechanically verifiable evidence satisfies predefined conditions. Crucially, the approach treats agent outputs as claims pending validation rather than established facts, and explicitly distinguishes between the mere existence of review mechanisms and their role as gating criteria. Empirical evaluation demonstrates zero false “DONE” declarations across ten scenarios, successful resistance against 18 classes of tampering attacks with no false acceptances, and a reduction in hidden failure rates from 31/1800 to 2/1800 in ablation studies. Furthermore, 94.8% of issues in a corpus of 565 self-application narratives were resolved.

This work addresses the problem of providing provably correct snapshot-equivalent change capture and replay for continuously written databases without relying on global snapshots or locking mechanisms. It introduces the “authenticated virtual slice” model, which combines interleaved log scanning with watermark-based positioning to enable incremental authentication over primary key ranges and advancing frontiers while preserving source continuity. For the first time, the paper formally defines and machine-verifies snapshot equivalence for DBLog, presenting a rigorous Isabelle/HOL proof that all well-formed executions satisfy per-key replay equivalence within their designated frontiers and key ranges, and that the authentication process yields valid virtual slices.

In cloud computing, untrusted storage interfaces are vulnerable to rollback and replay attacks, compromising the integrity of confidential application decisions. Existing hardware-enforced state continuity mechanisms treat all rollbacks as malicious, thus precluding legitimate rollbacks required for fault recovery. This paper introduces Rebound, the first framework enabling fine-grained differentiation between malicious and legitimate rollbacks in confidential cloud environments. Rebound features a policy-authorized reference monitor—hardware-protected via trusted execution—supporting atomic state updates, controlled rollback, and tamper-evident audit logging. Leveraging formal verification, policy-driven access control, and end-to-end logging, we evaluate Rebound in GitLab CI, demonstrating efficient and secure version management for binaries, configurations, and data. The system incurs low overhead while providing strong security guarantees against unauthorized or inconsistent state transitions.

In digital forensics, the atomicity and integrity of storage snapshots lack rigorous definitions that jointly guarantee both instantaneousness and causal ordering—undermining evidentiary admissibility in legal proceedings. To address this, we propose a novel atomicity definition grounded in causal consistency, overcoming the limitation of conventional time-based atomicity models. We further rectify conceptual flaws in existing integrity definitions and introduce a revised, theoretically sound yet engineering-practical integrity criterion—explicitly supporting copy-on-write (CoW) implementations. Our approach integrates causal modeling, formal snapshot semantics, CoW mechanism analysis, and formalization of forensic quality criteria, yielding a verifiable snapshot semantic framework. This work establishes the first theoretical foundation for forensic tool design that unifies causal ordering with instantaneous state capture, thereby significantly enhancing the forensic validity and judicial admissibility of live data acquisition.

Training large foundation models (LFMs) faces significant challenges in checkpoint management, including poor cross-framework compatibility, tight coupling with parallelization strategies, heterogeneous storage backends, and severe I/O bottlenecks. To address these, this work proposes an industrial-grade unified archival system. Its core contributions are: (1) a novel parallelism-agnostic checkpoint serialization format; (2) a full-stack I/O optimization framework integrating a dynamic resharding engine, multi-framework abstraction interfaces (PyTorch/Megatron/DeepSpeed), asynchronous high-throughput storage adapters, and a distributed I/O monitoring toolchain; and (3) runtime support for cross-parallelism resharding, multi-backend adaptivity, and rapid failure recovery. Experiments demonstrate an average 54.20× reduction in checkpoint blocking time, with peak checkpoint save and load speedups of 9.96× and 8.80×, respectively. The system has been stably deployed in production environments scaling to over one thousand GPUs.

This work addresses the substantial overhead in parallel proof search within Lean 4, which arises from repeatedly reconstructing proof states—such as library imports and theorem unfoldings—across different branches. To mitigate this redundancy, the authors propose a proof state snapshotting mechanism orthogonal to existing caching strategies. This approach captures and serializes the fully elaborated proof state once within the Lean 4 language server, enabling direct reuse across multiple search branches without recomputation. Integrated into the Snapshot-DSP inference pipeline, the technique achieves end-to-end speedups of 5.6× to 50× (averaging 14×) on the miniF2F-v2 benchmark, with performance gains scaling markedly as the number of parallel branches increases.

This work addresses the susceptibility of large language models to hallucinations in empirical reasoning—outputs lacking verifiable evidence or formal guarantees. The authors propose EG-VAR, an architecture that uniquely leverages the Lean 4 formal proof kernel as the sole trusted generator of claims. By integrating tool-certified axioms and a source elevation mechanism, EG-VAR ensures every output is bound to a kernel-verified chain of reasoning and tool invocation; otherwise, it abstains and provides a fully traceable audit trail. Evaluated on a TableBench subset, EG-VAR achieves perfect accuracy (120/120), substantially outperforming a 95% baseline. In counterfactual tests, it maintains 100% source fidelity—significantly higher than competing methods (80–90%)—and exhibits remarkably low semantic formalization error rates of 1.7% (Opus) and 3.3% (Sonnet).

This work addresses the challenge of determining whether a local recovery point is semantically valid when structured tool-using agents fail mid-execution, particularly in scenarios where downstream components have already committed to outputs from upstream stages. The paper introduces DART, a runtime system that formalizes the notion of “semantic recoverability” for the first time. DART enables safe and efficient partial recovery by identifying failure instances, verifying semantic boundaries, aligning checkpoints, and selecting legitimate recovery points under dependency and effect constraints. Its modular architecture incorporates explicit acceptability checks to prevent invalidation of already-committed downstream work. Empirical evaluation across three LLM-driven tasks and the LangGraph framework demonstrates that DART successfully recovers all commitment-sensitive cases where baseline methods fail, with no unsafe rollbacks detected in a five-domain safety audit.

Current large language model (LLM) agents lack verifiability, debuggability, and auditability, and relying solely on the accuracy of final answers fails to reveal their underlying reasoning. To address this, this work proposes the first unified provenance framework for LLM agents, systematically modeling causal relationships in tool usage, memory access, and environmental interactions. It introduces a comprehensive provenance taxonomy encompassing source, granularity, representation format, and trust functions. By integrating provenance-aware representation modeling, evidence attribution, runtime safeguards, provenance-informed memory management, and trajectory observability analysis, the study shifts the evaluation paradigm from outcome correctness to process accountability. The framework consolidates existing benchmarks to define a clear pathway for process-level trustworthiness assessment and highlights key challenges, including standardized trajectory schemas, semantic-level provenance, and privacy-preserving auditing.

This work addresses the breakdown of the graph isomorphism correctness theorem in the formal verification of the CertiGC garbage collector when extending from an immutable to a mutable environment, where the original assumption of “global absence of backward edges” no longer holds. To resolve this, the authors present the first integration of an AI agent (Codex) with the interactive theorem prover Rocq, leveraging the Verified Software Toolchain (VST) and CertiGraph frameworks to restructure the proof architecture. They introduce and formally verify a backward-edge invariant, thereby repairing the VST relational proof, restoring the graph isomorphism theorem, and eliminating erroneous assumptions between specification and theorem. This effort not only removes an outdated hypothesis but also yields a maintainable verification artifact, demonstrating the feasibility and effectiveness of large language models in assisting complex systems formal verification.