KAIST AI's FactEmbed decouples query understanding to improve composed image retrieval
kaist-ai · hf · 2026-11-04
KAIST AI released FactEmbed on Hugging Face, a new approach for composed image retrieval.
The key idea is decoupling query understanding from embedding construction: factorized latents simulate visual modifications on the reference image before retrieval, letting the model compose image semantics more faithfully to the text instruction.
More from Research
- 387 papers distill one question: when is a better model the answer in AI for science? — bravo_abad · 2026-11-04
- Noetra engineer's first project: building an image encoder from scratch for multimodal LLMs — HirokatuKataoka · 2026-11-04
- Japanese team trains image encoder from scratch to match SigLIP2, with better CJK performance — HirokatuKataoka · 2026-11-04
- SpecJudge author argues LLM-as-judge tools should never ask models for self-reported confidence — jokiruiz · 2026-11-04
- Tsinghua's BEMULA benchmark: no LLM agent memory system passes over 1.5% of forget requests — rohanpaul_ai · 2026-11-04
- Intology's Locus auto-designs molecular models, beating GPT-6 and Claude agents — rohanpaul_ai · 2026-11-04