§ feed · storyline
Aurora
Aurora releases an open-source reinforcement learning framework that converts speculative decoding into a self-improving system, reporting 1.25x gains over a well-trained static speculator.
1.25x over a well-trained static speculator. Aurora is an open-source RL framework that turns speculative decoding from a one-time offline setup into a self-improving system that learns from every request it serves.
§ sources1 publication · timeline below
- together.aiAuroraprimary