William Fleshman

SpectR was accepted at @COLM_conf! Our follow-up work, LoRA-Augmented Generation (LAG), combines SpectR w/ a first pass filtering of adapters using Arrow routing. LAG is significantly more efficient, enabling SpectR like performance with much larger LoRA libraries!

🚨Announcing TaCQ 🚨 a new mixed-precision quantization method that identifies critical weights to preserve. We integrate key ideas from circuit discovery, model editing, and input attribution to improve low-bit quant., w/ 96% 16-bit acc. at 3.1 avg bits (~6x compression)…

🚨 New preprint on transparent, tree-adaptive grounded reasoning! We introduce Bonsai, a versatile reasoning system that generates interpretable, grounded, and uncertainty-aware reasoning traces while enabling state-of-the-art performance on text and video benchmarks.

‼️Tired of dealing with long reasoning chains?‼️ Introducing Compressed Chain of Thought (CCoT), a framework to perform efficient reasoning through a few dense representations in place of long sequences of discrete tokens. 📜: arxiv.org/abs/2412.13171

William Fleshman brings it home with our final talk of #CAMLIS2024: "AdapterSwap: Continuous Training of LLMs with Data Removal and Access-Control Guarantees"!

Introducing ✨Promptriever ✨ the first retriever that can be prompted like an LM with free-form prompts! Our secret: query-level instruction training lets you keep the promptability of the base LM! 🚫 keyword-matching ✅ instruction search 📝 arxiv.org/abs/2409.11136

Thanks Ed!

Sakana AI is proud to sponsor the LLM Merging Competition: Building LLMs Efficiently through Merging at #NeurIPS2024 🤗 If you’re excited about pushing the frontiers of model merging, please visit: llm-merging.github.io