Retrieval researcher challenges 'lossless' claims for MRL and ANN search

Researcher srchvrs responds to debates on Matryoshka representation learning, arguing its loss is tolerable but that benchmarks like MS MARCO are too small, with outdated QRELs, making common claims that ANN retrieval is nearly lossless versus exact k-NN unreliable.

2026-10-26 ~ 2026-10-26 · 2 related posts

Full story(2 episodes)→

• Episode 1: Study Questions Value of Matryoshka Embeddings(2026-10-26, 3 posts)
• Episode 2: Retrieval researcher challenges 'lossless' claims for MRL and ANN search(2026-10-26, 2 posts)

• Search researcher on MRL debate: losses are tolerable but benchmarks are too small with stale QRELs — srchvrs · 2026-10-26
• ANN search isn't nearly lossless either: MS MARCO-style benchmarks are small and too easy — srchvrs · 2026-10-26