paper-with-me

홈 › Papers

Temporal Leakage in Search-Engine Date-Filtered Web Retrieval: A Retrospective Forecasting Case Study

2026-01-31 · Ali El Lahib, Ying-Jieh Xia, Zehan Li, Yuxuan Wang, Xinyu Pi arxiv

Search-engine date filters are widely used to enforce pre-cutoff retrieval in retrospective evaluations of search-augmented forecasters. We show this approach is unreliable across two major search engines: auditing Google Search's before: filter and DuckDuckGo's date-range filter, we find that at least one retrieved page contains major post-cutoff leakage for 71% of questions on Google and 81% on DuckDuckGo, and the answer is directly revealed for 41% and 55%, respectively. Using gpt-oss-120b to forecast with these leaky documents, we demonstrate inflated prediction accuracy (Brier score 0.10 vs. 0.24 with leak-free documents). We characterize recurring leakage mechanisms, including updated articles, related-content modules, unreliable metadata, and absence-based signals, and argue that date-restricted search on these engines is insufficient for credible retrospective evaluation. We recommend stronger retrieval safeguards or evaluation on frozen, time-stamped web snapshots.

📄 PDF Abstract BibTeX arXiv:2602.00758

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Scaling Point-in-Time Language Models

2026-04-24 · Bryan Kelly, Semyon Malamud, Johannes Schwab, Teng Andrea Xu arxiv

Large language models trained on unrestricted internet corpora inevitably embed information from the future, introducing lookahead bias that compromises the validity of backtests and causal inference in finance and the s…

LessLeak-Bench: A First Investigation of Data Leakage in LLMs Across 83 Software Engineering Benchmarks

2025-02-10 · Xin Zhou, Martin Weyssow, Ratnadira Widyasari, Ting Zhang 외

Large Language Models (LLMs) are widely utilized in software engineering (SE) tasks, such as code generation and automated program repair. However, their reliance on extensive and often undisclosed pre-training datasets …

Code GenerationProgram Repair

A Complete Year of User Retrieval Sessions in a Social Sciences Academic Search Engine

2017-09-23 · Mayr Philipp, Kacem Ameni

In this paper, we present an open data set extracted from the transaction log of the social sciences academic search engine sowiport. The data set includes a filtered set of 484,449 retrieval sessions which have been car…

Retrieval

Data Leakage in Automotive Perception: Practitioners' Insights

2026-04-08 · Md Abu Ahammed Babu, Sushant Kumar Pandey, Darko Durisic, Andras Balint 외 arxiv

Data leakage is the inadvertent transfer of information between training and evaluation datasets that poses a subtle, yet critical, risk to the reliability of machine learning (ML) models in safety-critical systems such …

Teaching Large Language Models When Not to Know: Learning Temporal Critique for Ex-Ante Reasoning

2026-05-14 · Chenlu Ding, Jiancan Wu, Yanchen Luo, Zheyuan Liu 외 arxiv

Large language models (LLMs) often fail to reason under temporal cutoffs: when prompted to answer from the standpoint of an earlier time, they exploit knowledge that became available only later. We study this failure thr…