TEVI: Sparse autoencoders for text-conditioned editing of CLIP image embeddings to improve vision-language alignment
TEVI is a framework that uses sparse autoencoders to disentangle CLIP image embeddings and a learned masking module to selectively reconstruct embeddings conditioned on a given caption, addressing the information imbalance between images and their captions. The approach improves image-text retrieval on both coarse-grained benchmarks (MS COCO, Flickr) and fine-grained long-caption benchmarks (IIW, DOCCI), with larger gains on richer captions. The work also shows improved robustness on the RoCOCO benchmark.
Related guides (2)
Related events (8)
Fine-Tuning CLIP with Remote Sensing Satellite Images and Captions
This Hugging Face blog post describes fine-tuning OpenAI's CLIP model on the RSICD (Remote Sensing Image Captioning Dataset) to improve vision-language alignment for satellite and aerial imagery. The work demonstrates domain adaptation of a general-purpose contrastive vision-language model to a specialized remote sensing context. It serves as a practical tutorial and case study for transfer learning with CLIP on narrow domains.
Hierarchical Text-Conditional Image Generation with CLIP Latents (DALL-E 2 / unCLIP)
OpenAI published research on hierarchical text-conditional image generation using CLIP latents, the technique underlying DALL-E 2. The approach uses a prior network to map text embeddings to image embeddings, then a diffusion decoder to generate images from those embeddings. This represented a significant advance in text-to-image generation quality and semantic fidelity at the time of release.
CLIP: Connecting Text and Images
OpenAI introduced CLIP (Contrastive Language-Image Pre-training), a neural network that learns visual concepts from natural language supervision. CLIP enables zero-shot visual classification by accepting natural language descriptions of categories rather than requiring task-specific training data. The approach mirrors the zero-shot transfer capabilities demonstrated by GPT-2 and GPT-3 in the language domain.
Chinese CLIP: Contrastive Vision-Language Pretraining in Chinese
Alibaba's Qwen team released Chinese CLIP, a language-specific vision-language contrastive pretraining model targeting Chinese multimodal representation learning. The project addresses a gap in open-source Chinese CLIP models, particularly for cross-modal retrieval tasks. It follows the CLIP framework but is adapted for Chinese language and cultural context.
ETCHR: Decoupled Image Editing for Visual Chain-of-Thought Reasoning in MLLMs
ETCHR introduces a question-conditioned, reasoning-aware image editing model that decouples visual transformation from downstream understanding in multimodal LLMs. It addresses two identified gaps—language-side (mapping abstract questions to visual edits) and generation-side (edit quality degrading with reasoning depth)—via a two-stage training recipe combining supervised fine-tuning on edit trajectories and VLM-derived reward signals. Because the editor is decoupled, it plugs into arbitrary MLLMs without retraining, yielding Pass@1 gains of roughly +4.6 to +5.5 points across five task families when paired with Qwen3-VL-8B, Gemini-3.1-Flash-Lite, and Kimi K2.5. The work advances the 'think with images' paradigm beyond fixed toolkits and unified multimodal approaches.
Zero-shot image-to-text generation with BLIP-2
Hugging Face published a blog post introducing BLIP-2, a multimodal model that enables zero-shot image-to-text generation by bridging frozen image encoders and large language models via a lightweight Querying Transformer (Q-Former). The post covers the model's architecture, capabilities, and how to use it via the Hugging Face Transformers library. BLIP-2 achieves strong performance on visual question answering and image captioning tasks without task-specific fine-tuning.
Training and Finetuning Sparse Embedding Models with Sentence Transformers
Hugging Face published a tutorial on training and fine-tuning sparse embedding models using the Sentence Transformers library. Sparse embeddings offer an alternative to dense vector representations for retrieval tasks, potentially improving interpretability and efficiency. The post covers the tooling and workflows available in Sentence Transformers for producing sparse encoders suitable for search and RAG pipelines.
BabyCL: Continual multimodal learning from egocentric child video in a single chronological pass
Researchers introduce BabyCL, a continual learning framework that processes the SAYCam egocentric child video dataset in a single chronological pass rather than shuffled multi-epoch training, more closely mimicking how children actually encounter their environment. The system combines streaming visual representation learning with image-text contrastive objectives, a multi-stage temporal segmentation, and a dual replay buffer managing visual and multimodal histories. BabyCL outperforms streaming baselines on the SAYCam Labeled-S 4AFC benchmark under matched compute budgets, substantially closing the gap to offline training upper bounds. The work advances understanding of whether neural networks can acquire word-referent mappings under biologically plausible training conditions.