Data-Centric AI Governance: Addressing the Limitations of Model-Focused Policies
Not on the map. Why?
Abstract (arXiv)
Current regulations on powerful AI capabilities are narrowly focused on "foundation" or "frontier" models. However, these terms are vague and inconsistently defined, leading to an unstable foundation for governance efforts. Critically, policy debates often fail to consider the data used with these models, despite the clear link between data and model performance. Even (relatively) "small" models that fall outside the typical definitions of foundation and frontier models can achieve equivalent outcomes when exposed to sufficiently specific datasets. In this work, we illustrate the importance of considering dataset size and content as essential factors in assessing the risks posed by models both today and in the future. More broadly, we emphasize the risk posed by over-regulating reactively and provide a path towards careful, quantitative evaluation of capabilities that can lead to a simplified regulatory environment.
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Builds on 53 of its 89 references that are in the map. First posted on arXiv 25 Sep 2024.
Builds on
53 of 89 references.
References · 53
- 2020Language Models are Few-Shot Learnerscited by 32,199
- 2022Training language models to follow instructions with human feedbackcited by 15,712
- 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,160
- 2020Measuring Massive Multitask Language Understandingcited by 7,768
- 2020Scaling Laws for Neural Language Modelscited by 6,411
- 2023Toolformer: Language Models Can Teach Themselves to Use Toolscited by 4,072
- 2022Constitutional AI: Harmlessness from AI Feedbackcited by 2,950
- 2022Training Compute-Optimal Large Language Modelscited by 2,927
- 2015Neural Machine Translation of Rare Words with Subword Unitscited by 2,898
- 2022Emergent Abilities of Large Language Modelscited by 2,483
- 2023Retrieval-Augmented Generation for Large Language Models: A Surveycited by 2,185
- 2023Jailbroken: How Does LLM Safety Training Fail?cited by 1,565
38 more
- 2024Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Dualitycited by 1,283
- 2023Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversationscited by 1,124
- 2018Model Cards for Model Reportingcited by 1,068
- 2021Scaling Vision Transformerscited by 993
- 2022Red Teaming Language Models with Language Modelscited by 983
- 2022Quantifying Memorization Across Neural Language Modelscited by 755
- 2022Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learnedcited by 702
- 2023The Falcon Series of Open Language Modelscited by 526
- 2020Dataset Condensation with Gradient Matchingcited by 510
- 2020What Neural Networks Memorize and Why: Discovering the Long Tail via Influence Estimationcited by 492
- 2019Does Learning Require Memorization? A Short Tale about a Long Tailcited by 452
- 2023TinyStories: How Small Can Language Models Be and Still Speak Coherent English?cited by 425
- 2023Scalable Extraction of Training Data from (Production) Language Modelscited by 415
- 2023The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A"cited by 399
- 2024Can LLMs Generate Novel Research Ideas? A Large-Scale Human Study with 100+ NLP Researcherscited by 348
- 2021AI and the Everything in the Whole Wide World Benchmarkcited by 340
- 2024RAFT: Adapting Language Model to Domain Specific RAGcited by 258
- 2023T2I-CompBench++: An Enhanced and Comprehensive Benchmark for Compositional Text-to-image Generationcited by 253
- 2024A Survey on Data Selection for Language Modelscited by 230
- 2023Sequential Modeling Enables Scalable Learning for Large Vision Modelscited by 212
- 2024World Model on Million-Length Video And Language With Blockwise RingAttentioncited by 196
- 2023Universal Instance Perception as Object Discovery and Retrievalcited by 190
- 2024Audio Flamingo: A Novel Audio Language Model with Few-Shot Learning and Dialogue Abilitiescited by 183
- 2022Data Cards: Purposeful and Transparent Dataset Documentation for Responsible AIcited by 169
- 2022Can Machines Help Us Answering Question 16 in Datasheets, and In Turn Reflecting on Inappropriate Content?cited by 150
- 2019xBD: A Dataset for Assessing Building Damage from Satellite Imagerycited by 138
- 2024Observational Scaling Laws and the Predictability of Language Model Performancecited by 128
- 2022Retrieval-Augmented Multimodal Language Modelingcited by 125
- 2023Sudden Drops in the Loss: Syntax Acquisition, Phase Transitions, and Simplicity Bias in MLMscited by 121
- 2023Frontier AI Regulation: Managing Emerging Risks to Public Safetycited by 95
- 2023Identifying and Mitigating the Security Risks of Generative AIcited by 90
- 2023TinyGPT-V: Efficient Multimodal Large Language Model via Small Backbonescited by 77
- 2021Pixelated Butterfly: Simple and Efficient Sparse training for Neural Network Modelscited by 75
- 2023Attack Prompt Generation for Red Teaming and Defending Large Language Modelscited by 73
- 2023The Data Provenance Initiative: A Large Scale Audit of Dataset Licensing & Attribution in AIcited by 73
- 2024Linguistic Bias in ChatGPT: Language Models Reinforce Dialect Discriminationcited by 55
- 2023Hierarchical Open-vocabulary Universal Image Segmentationcited by 53
- 2023Open, Closed, or Small Language Models for Text Classification?cited by 34
50 most cited of 53 All 53 (JSON)
Built on by
Citing papers per year: 2025: 1; 2026: 1.
2 papers.