ScalarLens splits coordinate from context in numerical embeddings, winning 25 of 27 CTR setups
_reachsumit · x · 2026-09-25
ScalarLens introduces a numerical embedding for CTR prediction that separates where a value lies from what it means for the current sample: a monotone local mesh builds a stable coordinate from the scalar alone, while bounded low-rank dynamics produce a contextual response—no normalization required.
- Motivation: on Criteo, the same numerical interval carries residual click evidence with opposite signs across contexts, and external normalization requires train/serve synchronization
- Evaluated in 1,539 runs across 19 representations, 3 datasets, 9 backbones, and 3 seeds
- Ranks first in 25 of 27 settings, second in the rest; ablations show scale correction or extra capacity alone don't reproduce the gain
More from Research
- Protein design team secures tons of GPUs and lab equipment in single-day buildout — nlarusstone · 2026-09-25
- FD-Loss Achieves 0.75 One-Step Pixel-Space FID, Accepted as NeurIPS Oral — yuewang314 · 2026-09-25
- The 'No Magic OOD' Principle: Flagging a Common Flaw in Alignment Plans — nabla_theta · 2026-09-25
- NeurIPS 2026 paper shows noise-conditioning-free diffusion models implicitly implement Riemannian gradient flow — docmilanfar · 2026-09-25
- Solving the AI Science Validation Bottleneck: Why Dario's Gene-Editing Claim Isn't the Whole Story — Ghost_Pilot_MD · 2026-09-25
- AI PhD Laments: ICLR Submissions Are Formulaic While Industry Labs Sprint Ahead — vykthur · 2026-09-25