物書きがLLMの頭の中を覗こうと頑張る連載です。
文学者は金門橋を渡ることができるか①｜浅間香織 KaoriAsama note.com/kaori_asama/...
#AI #LLM #生成AI #SAE #sparseautoencoder #goldengatebridge
2
0
0
0
#sparseautoencoder
Latest posts tagged with #sparseautoencoder on Bluesky
Trending
Posts tagged #sparseautoencoder
SAEdit: Token‑Level, Continuous Image Editing with Sparse Autoencoder
SAEdit lets users edit images by shifting token embeddings along autoencoder directions, keeping other visuals unchanged. Released on 6 Oct 2025 (arXiv:2510.05081). getnews.me/saedit-token-level-conti... #saedit #sparseautoencoder #diffusion
0
0
0
0
Superposition Disentanglement Improves Neural Representation Alignment
Sparse autoencoders disentangled superposition, boosting alignment scores between visual AI models and brain recordings, the study notes. October 2025. Read more: getnews.me/superposition-disentangl... #superposition #sparseautoencoder
0
0
0
0
Refining LLM Steering Vectors with Sparse Autoencoders
Researchers use sparse autoencoders to refine LLM steering vectors via denoising and augmentation, improving performance on limited data. Submitted 28 Sep 2025 (arXiv:2509.23799). Read more: getnews.me/refining-llm-steering-ve... #llmsteering #sparseautoencoder
0
0
0
0
New Study Introduces Metric for Sparse Autoencoder Feature Sensitivity
A new 'feature sensitivity' metric for sparse autoencoders, measuring activation stability, was submitted in September 2025 to the NeurIPS 2025 Workshop. getnews.me/new-study-introduces-met... #featuresensitivity #sparseautoencoder
0
0
0
0
Sparse Autoencoders Help Interpret Large Multimodal Models
Researchers used a Sparse Autoencoder on LLaVA‑NeXT‑8B and leveraged LLaVA‑OV‑72B to auto‑generate natural‑language descriptions of each feature, improving empathy tasks. Read more: getnews.me/sparse-autoencoders-help... #sparseautoencoder #llava
0
0
0
0