Share your thoughts, 1 month free Claude Pro on usSee more

Conflict Resolution and Memory Verification on MemoryAgentBench (conflict-bearing slice)

100Accuracy

Oracle

Updated 18d ago

Evaluation Results

Method	Links
Oracle 2026.07		100	100	0
Collapsed 2026.07		97.5	0	2.5
StateFuse (cons.) 2026.07		64.9	100	2.1
StateFuse core 2026.07		64.9	100	2.1
MV register (cons.) 2026.07		64.9	100	2.1
Provenance surface (cons.) 2026.07		64.9	100	2.1
Raw log (cons.) 2026.07		64.9	0	2.1