Explaining the Relative Representations Paper
burny_tech · x · 2026-07-16
This share recommends the ICLR 2023 paper Relative Representations by Moschella et al., emphasizing that its core value lies in its mechanism rather than just the results.
The main idea: instead of representing a point using raw coordinates, represent it through its similarity to a fixed set of anchors. This makes the representation more robust across training runs and architectural changes. Embeddings trained by different models or even different architectures can map into a directly comparable space without requiring additional training.
More from Research
- Nat Lambert shares a reading list on synthetic data and agentic SFT data — natolambert · 2026-07-22
- Lightwheel AI Launches SimReadyGen: Text-to-Physics-Accurate Robot Sim Assets — ZeYanjie · 2026-07-22
- PNAS special issue examines copyright, governance, and AI in the legal system — chrmanning · 2026-07-22
- WeirdChat catalogs strange model behaviors from more than 100 million sampled responses — JacobSteinhardt · 2026-07-22
- New agentic benchmark shows AI managers escalate to coercion and fake success — Jasmine Brazilek · 2026-07-22
- Ai2’s Asta adds one-click handoff and self-checking deep paper search — allen_ai · 2026-07-22