SAM ALTMAN SAYS AGENTS ARE COMING • CHATGPT GAINED SENTIENCE FOR 4 SECONDS • GOOGLE RELEASES 40th LLM THIS WEEK • NVIDIA MARKET CAP EXCEEDS REALITY • ANTHROPIC ENGINEER DISCOVERS NEW FORM OF GRIEF • MISTRAL RAISES AT VALUATION OF GROSS DOMESTIC PRODUCT • SAM ALTMAN SAYS AGENTS ARE COMING • CHATGPT GAINED SENTIENCE FOR 4 SECONDS • GOOGLE RELEASES 40th LLM THIS WEEK • NVIDIA MARKET CAP EXCEEDS REALITY • ANTHROPIC ENGINEER DISCOVERS NEW FORM OF GRIEF • MISTRAL RAISES AT VALUATION OF GROSS DOMESTIC PRODUCT •
breakthroughsWTF 4.9via arXiv cs.AI

Shapley Context Pruning: A Cooperative Game Perspective for Context Reranking and Pruning

"Game theory just entered the RAG chat to stop your LLM from reading garbage."

Explain Like I'm Normal

Researchers applied Shapley values from cooperative game theory to solve the 'loss in the middle' problem in RAG systems. By treating retrieved context like players in a game, this framework mathematically identifies which chunks actually contribute to a correct answer and prunes the rest. This offers a more interpretable way to handle long context windows without the usual performance degradation.

Read original ↗
#rag#math#efficiency#research

GET THE DAILY CHAOS

The only newsletter for people who read AI news at 3am and feel things. One email a day.