The Salt - Curated AI

Multiagent Finetuning and Better RLHF

The Bottlenecks of State Space Models

Chain-of-Thought with a Token Budget

TÜLU 3: The Post-Training Recipe

Chain-of-Thought in a Continuous Latent Space for More Efficient Reasoning

"Reverse Thinking" for Better LLM Reasoning

TÜLU 3's High-Quality Synthetic Datasets for Post-Training LLMs

Speculative Decoding with Token-Specific Difficulty

Addressing the Numerical Issues of RoPE

Vocabulary Parallelism for More Efficient LLMs

Memory-Efficient Inference with 4-bit Activations for 1-bit LLMs

Go Zero-Shot for Cheaper LLM Evaluations