Three Levers, One Loop

Ai2 calls it the model flow: the pretrained base, the mid-training checkpoints, the supervised fine-tuning stage, the DPO stage, the reinforcement learning stage, the training code, the data recipe, and the datasets themselves, all…

Checkpoints from each training stage are published alongside the final model (Hugging Face · Olmo 3.1 32B Instruct).That level of openness is the precondition for everything below.

Coding costs cut by as much as 56 percent with a 2 percent decline in quality. Mark Austin, AT&T vice president for employee AI tools, as reported by PYMNTS, August 20, 2026 The industry is moving in this direction for reasons that are not…

It freezes the base weights and trains small low-rank adapters, cutting trainable parameters by orders of magnitude while keeping most of the quality of full fine-tuning.

It replaces the reward model and the PPO loop with a single closed-form loss over chosen-versus-rejected pairs (Rafailov et al. · DPO, NeurIPS 2023). Eighteen rounds against the golden set Pass rate per round (bars) and the promotion bar…

Controlled studies of fine-tuning find significant variance from the random seed alone, at the level of aggregate metrics and at the level of individual predictions (Bui, Savova & Wang · IJCNLP-AACL 2025). ### Below a threshold data scale,…

Its authors concluded that almost all knowledge in large language models is learned during pretraining, and only limited instruction tuning data is necessary to teach models to produce high quality output.

Its recipe treats a standardized evaluation suite for development, decontamination, and final evaluation as a first-class stage alongside data curation, SFT, DPO, and RL (Lambert et al. · Tülu 3).Four evaluation decisions changed how our…

It estimates the advantage from a group of sampled completions instead of a learned value function, the same optimizer behind DeepSeek's R1 reasoning training (Shao et al. · DeepSeekMath (GRPO)).