arXiv AI

Trust, but verify, audit and attribute: a human-in-the-loop framework for checking AI-produced research results

The paper proposes a human‑in‑the‑loop framework for verifying AI‑generated research results, particularly in mathematics and statistics. It outlines a system where AI agents perform checks using specialized software, a fidelity audit ensures each check addresses the claim, and humans intervene at fixed points to decide on outcomes. The framework also records contributions and attributes work to both human and AI, and the authors demonstrate two automatic implementations for mathematics and statistics, including a case study where an ambiguous claim was split into two readings, one refuted and one proved.

Hugging Face Trending Papers
Nov 9

Trust, but verify, audit and attribute: a human-in-the-loop framework for checking AI-produced research results

The paper proposes a human‑in‑the‑loop framework for verifying AI‑generated research results, addressing the growing volume of AI‑produced mathematics manuscripts. It outlines a system where AI agents perform checks using specialised software, but a human researcher remains involved to audit fidelity, decide at fixed points, and attribute contributions. The authors present two automatic implementations—one for mathematics (using Wolfram Language, MATLAB, Lean) and one for statistics (Python and R)—and demonstrate the mathematical tool on a two‑claim example that required splitting an ambiguous claim into a refuted and a proved reading.

arXiv AI
Aug 25

CausalSmith: A Formally Grounded, Self-Improving Agentic Framework for Automated Research in Causal Inference

CausalSmith is a framework that automates theoretical research in causal inference by integrating a Lean proof assistant with a self‑improving agentic pipeline. It uses Causalean, a Lean library of over 7,000 machine‑checked declarations, and a pipeline that selects topics, proposes results, formalizes statements, constructs proofs, and audits them against informal claims. The system’s artifacts and source code are publicly available on GitHub.

By Jiyuan Tan, Vasilis Syrgkanis
arXiv AI
Oct 31

Beyond fabrication: Misattributed numbers in papers written by AI research agents

The paper examines how AI‑written research papers often contain misattributed numerical claims. By re‑analyzing 540 judgments from a benchmark of nine AI agents on fifteen research problems, the authors manually labeled 256 claims and found that most failures stem from misalignment between the paper and the underlying work rather than outright cheating. They propose a deterministic claim‑to‑evidence checker that achieves over 90% precision and, when combined with an LLM verifier, offers a more reliable audit of AI‑generated research.

By Mosab Alfaqeeh, Omar Hourani, Muath Hourani, Hashem Alfaqeeh
arXiv AI
Jul 17

MathCoPilot: An Interactive System for Human-AI Symbiotic Paradigm of Mathematical Research

arXiv:2607. 14582v1 Announce Type: new Abstract: Existing LLM-based theorem provers have achieved impressive results on formal mathematics benchmarks, yet they remain confined to acting as autonomous agents that prove a stated proposition.

By Junjie Zhang, Jiayu Liu, Wenbin Liu, Zhenya Huang, Doudou Wang, Yan Jiang, Leiye Xu, Tao Xiong, Wen Huang, Qi Liu, Guoping Hu, Enhong Chen, Mengping Zhang, Xiangdong Ye
arXiv AI
Oct 27

KAPPA: Checking the Proofs in Statistics Papers with Lean 4 and Language Model Agents

KAPPA is an open‑source toolkit that uses language‑model agents to translate LaTeX proofs from statistical papers into Lean 4, where each numbered result is verified step by step against the original text. The system assigns one of four verdicts—verified unconditionally, verified conditionally, issue recorded, or unfinished—to every result, and it can run unattended, re‑check only changed sections, and report resource usage. In tests on four example papers and three published statistical papers, KAPPA detected the known errors and additional issues that were later confirmed by hand, while a direct language‑model approach without the kit produced fewer correct verdicts and sometimes over‑claimed proofs.

By Zihua She, Xiao Wang
arXiv AI
Jun 16

SorryDB: Can AI Provers Complete Real-World Lean Theorems?

arXiv:2603. 02668v2 Announce Type: replace Abstract: We present SorryDB, a dynamically-updating benchmark of open Lean tasks drawn from 78 real world formalization projects on GitHub.

By Austin Letson, Leopoldo Sarra, Auguste Poiroux, Oliver Dressler, Paul Lezeau, Dhyan Aranha, Frederick Pu, Aaron Hill, Miguel Corredera Hidalgo, Julian Berman, George Tsoukalas, Lenny Taelman
arXiv Computation and Language
Sep 22

Checkpoints Are Not Enough: Trust Calibration in CoSLR, a Human-AI System for Systematic Literature Reviews

The paper introduces CoSLR, a Human‑AI collaborative system for systematic literature reviews that incorporates mandatory human checkpoints within a three‑phase pipeline using large language models and Retrieval‑Augmented Generation. In a survey of 63 participants, 42.9 % rated the system’s usability highly, yet 34.9 % indicated they would trust AI‑generated summaries without further human verification after brief interaction. The study highlights that effective human oversight in AI‑assisted literature reviews depends on users’ willingness to engage with the checkpoints, underscoring a calibration issue that interface design must directly address.

By MD Aidul Islam, Malik Abdul Sami, Muhammad Waseem, Zeeshan Rasheed, Kai-kristian Kemell, Zheying Zhang, Pekka Abrahamsson
arXiv AI
Oct 2

FORALL-LEAN-AGENT for Auditable Reasoning in Formal Mathematics and Software Verification

FORALL-LEAN-AGENT is a frontend‑agnostic framework that enhances auditable reasoning in formal mathematics and software verification. It integrates isolated workspaces, Lean tools, and fresh review processes—including statement comparison, axiom audits, and independent proof checking—to bind verification evidence and reviewer decisions to the same candidate artifact, ensuring traceable acceptance. Evaluations on VeriSoftBench, PutnamBench, and the Lean Eval track show that the framework improves benchmark success rates and reduces costs while providing evidence beyond aggregate solve counts.

By Naing Oo Lwin