OptST adapts frozen semantic tokenizers to downstream LLM recommenders via low-rank deformation
_reachsumit · x · 2026-11-06
OptST makes a pretrained semantic tokenizer responsive to downstream recommendation. It learns a low-rank, anchor-regularized deformation of token assignments without updating the pretrained encoder or base codebooks, trains through a straight-through estimator while keeping hard nearest-code assignments at inference, and selects an interaction-conditioned prefix depth to control semantic resolution per interaction.
More from Research
- UWaterloo 3D prints Invar36 flexures with directly mounted interferometer optics — jwt0625 · 2026-11-06
- AMD paper: a single generalized MLP upscales textures on-the-fly at render time, saving VRAM and disk — ssh4net · 2026-11-06
- OpenAI researcher invokes 2019 double descent paper in scaling-law spat — willdepue · 2026-11-06
- Encord debuts REACH egocentric hand-tracking model, cutting errors up to 70% — Scobleizer · 2026-11-06
- Only 16% of AI benchmarks measure organizational difficulties of real jobs — ben_burtenshaw · 2026-11-06
- Skillware framework claims 97.3% cost cut as 3B-8B local models beat frontier LLMs — RossPeili · 2026-11-06