Alibaba's PailitaoMMSearch 2.0: one VLM maps image+query straight to product IDs
_reachsumit · x · 2026-10-29
Alibaba's technical report presents PailitaoMMSearch2.0, a single vision-language model that directly generates product Semantic IDs from a raw query image and free-form instruction for unified end-to-end multimodal search. It scales along three dimensions: data scaling via a reusable toolset for cold-start synthesis and continuous data production; intent scaling via continual pre-training, mixed-budget SFT and unified RL, giving one model multiple retrieval relations and adjustable reasoning budgets; and adaptation to online scenarios with different resource constraints—decoupling search from fixed scenarios.
More from Research
- Lie detector compiled into Qwen3.5 weights flags 91% of lies with zero false positives — teortaxesTex · 2026-10-29
- Practitioner: No serious lab actually trains on the test set today — nrehiew_ · 2026-10-29
- Researchers slam secrecy around OpenAI-linked Yang-Mills proof materials — anshulkundaje · 2026-10-29
- KITP forms Yang-Mills Working Group to vet AI-related proof, drawing skepticism — anshulkundaje · 2026-10-29
- Job-market paper: German engineering societies lifted regional patents 4.8% — Afinetheorem · 2026-10-29
- François Fleuret Skeptical of Self-Improvement Training: Only Boosts the Easy Early Part — francoisfleuret · 2026-10-29