New research quantifies agreement between data-influence and data-similarity in LLMs

Researchers have quantified the agreement between data-similarity and data-influence measures used to trace LLM outputs back to their training data. Their findings indicate a significant overlap between the two measures, with data-influence assigning more consistent ranks to the top documents identified by data-similarity. This asymmetry was observed across experiments with models including OLMo2-1B, Qwen3-1.7B, LlaMa3.2-1B, Gemma3-1B, and GPT2. The study proposes leveraging this asymmetry to achieve a better cost-accuracy trade-off by using data-influence to refine data-similarity results. AI

IMPACT Provides a new method for understanding LLM behavior and potentially optimizing training data analysis.

RANK_REASON The cluster contains an academic paper detailing a new research methodology for understanding LLM behavior. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• Christopher Anderson
• Gemma3-1B
• GPT2
• LlaMa3.2-1B
• OLMo2-1B
• Qwen3-1.7B

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →