paper-with-me

홈 › Papers

URLs Help, Topics Guide: Understanding Metadata Utility in LLM Training

2025-05-22 · Dongyang Fan, Vinko Sabolčec, Martin Jaggi

Large Language Models (LLMs) are commonly pretrained on vast corpora of text without utilizing contextual metadata such as source, quality, or topic, leading to a context-free learning paradigm. While recent studies suggest that adding metadata like URL information as context (i.e., auxiliary inputs not used in the loss calculation) can improve training efficiency and downstream performance, they offer limited understanding of which types of metadata are truly effective and under what conditions. In this work, we conduct a systematic evaluation and find that not all metadata types contribute equally. Only URL context speeds up training, whereas quality scores and topic/format domain information offer no clear benefit. Furthermore, the improved downstream performances of URL conditioning emerge only when longer prompts are used at inference time. In addition, we demonstrate that context-aware pretraining enables more controllable generation than context-free pretraining, in a classifier-free guidance fashion. Although topic and format metadata do not accelerate training, they are effective for steering outputs, offering human-interpretable control over generation.

📄 PDF Abstract BibTeX arXiv:2505.16570

Code (0)

등록된 구현이 없습니다.

Methods 이 논문이 사용한 방법론

URL A computationally efficient approach for solving hard nonlinear problems of reinforcement learning (RL). It combines umbrella sampling, from computational physics/chemistry, with…

Similar Papers 제목 키워드 기반

Beyond URLs: Metadata Diversity and Position for Efficient LLM Pretraining

2025-11-26 · Dongyang Fan, Diba Hashemi, Sai Praneeth Karimireddy, Martin Jaggi arxiv

Incorporating metadata in Large Language Models (LLMs) pretraining has recently emerged as a promising approach to accelerate training. However prior work highlighted only one useful signal-URLs, leaving open the questio…

Metadata Conditioning Accelerates Language Model Pre-training

2025-01-03 · Tianyu Gao, Alexander Wettig, Luxi He, Yihe Dong 외

The vast diversity of styles, domains, and quality levels present in language model pre-training corpora is essential in developing general model capabilities, but efficiently learning and deploying the correct behaviors…

Language ModelingLanguage Modellingmodel

Uncovering Semantics and Topics Utilized by Threat Actors to Deliver Malicious Attachments and URLs

2024-07-11 · Andrey Yakymovych, Abhishek Singh

Recent threat reports highlight that email remains the top vector for delivering malware to endpoints. Despite these statistics, detecting malicious email attachments and URLs often neglects semantic cues linguistic feat…

4kClusteringDiversitySemantic Similarity+1

Automatic Generation of Web Censorship Probe Lists

2024-07-11 · Jenny Tang, Leo Alvarez, Arjun Brar, Nguyen Phong Hoang 외

Domain probe lists--used to determine which URLs to probe for Web censorship--play a critical role in Internet censorship measurement studies. Indeed, the size and accuracy of the domain probe list limits the set of cens…

Keyword Extraction

Characterizing Malicious URL Campaigns

2021-08-29 · Mahathir Almashor, Ejaz Ahmed, Benjamin Pick, Sharif Abuadbba 외

URLs are central to a myriad of cyber-security threats, from phishing to the distribution of malware. Their inherent ease of use and familiarity is continuously abused by attackers to evade defences and deceive end-users…