Good point.
This would be a clean way to test actual belief revision. LongMemEval and MemoryAgentBench explore related problems, but your paired-query setup could isolate this more directly. Thanks for the suggestion.
Rahul Vimalkanth PRO
boltwave
AI & ML interests
Embodied AI
Recent Activity
commentedon their article 21 days ago
The Missing Piece in LLM Memory: Why Long Context and Retrieval Aren't Enough for Belief Revision published an article 23 days ago
The Missing Piece in LLM Memory: Why Long Context and Retrieval Aren't Enough for Belief RevisionOrganizations
commented on The Missing Piece in LLM Memory: Why Long Context and Retrieval Aren't Enough for Belief Revision 21 days ago
published an article 23 days ago
Article
The Missing Piece in LLM Memory: Why Long Context and Retrieval Aren't Enough for Belief Revision
boltwave
• • 62