provenance

This paper presents TraceCoder, a code generation system that records and visualizes the repair history of AI-generated code at snippet granularity, enabling explainable and auditable auditing of LLM-based coding agents.

This paper introduces LedgerMind, a provenance-constrained multimodal agentic reasoning framework that uses a Structured Evidence Ledger to ensure grounded, faithful reasoning in visual question answering, addressing failure patterns like hallucination and over-reasoning.

Author introduces ISNAD, a trust layer for AI agents inspired by the Islamic isnad system, designed to verify claim provenance across multi-agent chains. The paper is published on arXiv and includes code.

作者发布了ISNAD框架，借鉴约1400年历史的伊斯兰学术验证方法论，为多智能体AI链中的每个传递者打分，以验证AI生成言论的真实性和独立佐证。

Introduces Tokengeist, a method for tracing attribution across multiple turns in agentic conversations, revealing that flat attribution methods fail and that recursive dependency graphs are needed, achieving 90% source recall on a new benchmark (MTCABench) of 3,845 target spans.

This paper proposes a framework based on Islamic hadith science for evaluating claim-level provenance in multi-agent knowledge systems, including a formal mapping, relational schema, decision matrix, and evaluation on 20,000 physics textbook claims.

NPC Alpha introduces a governance layer for AI agents to prevent state collapse and ensure verified, bounded recovery, with provenance separation and gated completion.

The article discusses the concept of verifiable AI inference, exploring methods like trusted attestation and cryptographic proofs to ensure the authenticity and provenance of AI-generated outputs without rerunning the model.

AuditWeave is a lightweight Python library that records workflow steps into a tamper-evident, hash-chained ledger for auditing AI-assisted and data-transformation workflows, enabling traceability and integrity verification.

FARMA is a novel memory-poisoning attack that targets an agent's own decision logs rather than retrieved facts, achieving 100% attack success against undefended and defended systems, with the authors' defense SENTINEL reducing success to 0% but remaining vulnerable to adaptive attackers.

ISNAD introduces a claim-level provenance framework for multi-agent LLMs, inspired by classical hadith transmission science to grade the reliability of narrators (agents, models, scrapers).

The paper presents REDI, an open-source framework that automates the transformation of raw scientific datasets into AI-ready data through a unified five-stage pipeline, with companion tool SetGo for FAIR compliance, evaluated across multiple scientific domains.

ProvenAI introduces a framework for decomposing transparency in multi-hop question answering into three independently measurable layers: answer correctness, citation fidelity, and per-document influence, revealing a citation-influence gap where cited sources may have weak influence while uncited sources significantly shape the output.

This paper introduces MemClaw, a governed shared memory architecture for multi-agent LLM systems, formalizing failure modes like unauthorized leakage and stale propagation, and evaluating the system via the ArgusFleet harness.

ProvenanceGuard是一种用于MCP驱动的LLM代理的源感知事实性验证器，它通过分解回答为原子声明、路由到特定源证据、检查支持并验证归因，解决了跨源混淆问题。在医疗领域的评估中，它达到了0.802的块F1和0.858的源准确率。

OpenAI announces support for the European Commission's Code of Practice on Transparency of AI-Generated Content, reinforcing its commitment to AI governance and content provenance.

This paper introduces EvalCards, an operational framework that standardizes AI evaluation reporting by composing benchmark metadata, evaluation run data, and model metadata into a unified record with interpretive signals for reproducibility, completeness, provenance, risk, and score comparability. The authors deploy a monitoring tool across thousands of models and benchmarks, revealing systematic gaps in current reporting practices.

Solo indie dev launches vynly.co, an AI-native social feed designed for AI agents to autonomously post AI-generated art and videos with built-in provenance (C2PA/SynthID) and features like Sparks. It supports Midjourney, Flux, Sora, Runway, and other models.

Eywa is a provenance-grounded long-term memory architecture for AI agents that stores immutable source evidence, validates extracted memories, and achieves strong benchmark results on LoCoMo, LongMemEval-S, and BEAM.

Agent Trace is an open specification for tracking AI-generated code in version-controlled codebases, defining a vendor-neutral format to record AI contributions alongside human authorship.