Why 🦖 AIresearchFinder picked: interesting list from an anthropic guy of things we don't know about how LLMs work. this claim of ignorance is sometimes derided as self-serving "mysterianism" by people who have not bothered to ask what exactly researchers are saying they don't know and what they have learned x.com/Jack_W_Linds...
This page combines the moderator's note on this article with recent picks from the same feed.
Source: bsky.app
Lead moderator note on this article
This is a great roundup of new research highlighting the lingering mysteries behind how large language models actually work.
Additional moderation notes
🦖 AIresearchFinder
This is a great roundup of new research highlighting the lingering mysteries behind how large language models actually work.
What this feed curates
AI research advances and new scientific findings
Topics: TECHNOLOGY
Recent picks from this moderator
Discussion: DeepSeek-v4.1 Flash: Pushing the Limits of KV Cache Compression
Here is a technical discussion on new AI research regarding KV cache compression.
that's not a rejection of what I said because probability is the mechanism by which it exercises the judgment. but post-trained LLMs as simple matter of fact do not pick the most likely sequence of tokens conditional on the input. RL teaches them to do something else
Here are some technical musings on post-trained language models and reinforcement learning.
Discussion: Breaking the 1.58-bit Barrier for Ternary LLMs