analiza aiqbench
This paper presents a fully open and reproducible post-training recipe for GLM-4.5-Air, detailing an eight-stage pipeline that transitions from supervised fine-tuning to advanced agent capabilities using public data without new human annotation. The study demonstrates that rigorous difficulty filtering and reliable reward design are critical for maintaining productive learning ranges while achieving performance competitive with official releases.
Rufus-Air: An Open LLM Post-Training Recipe
arXiv:2609.29421v1 Announce Type: cross Abstract: Rufus-Air is an open and reproducible post-training recipe on GLM-4.5-Air-Base (106B-A12B), organized as a serial pipeline of eight stages: SFT, Reasoning RL, Coding RL, Instruction-Following RL, General Agent, Coding Agent, Search Agent, and RLHF. We document the data, reward design, infrastructure, stage order, and stagewise results needed to reproduce the recipe. Stages progress from basic to advanced capabilities and from hard, verifiable rewards to softer judge-based signals. Training builds on open-source components and public data, much of it used as released, without new human annotation or an in-house distillation teacher. Our main findings are that (i) diverse, high-quality SFT establishes a strong capability floor; (ii) difficulty filtering keeps RL prompts within a productive learning range; (iii) reward reliability provides a practical principle for ordering stages; and (iv) infrastructure and engineering choices are part of the recipe, not just an implementation detail. Rufus-Air improves over the official GLM-4.5-Air post-trained release and is competitive with similarly sized open models.
Abstract:Rufus-Air is an open and reproducible post-training recipe on GLM-4.5-Air-Base (106B-A12B), organized as a serial pipeline of eight stages: SFT, Reasoning RL, Coding RL, Instruction-Following RL, General Agent, Coding Agent, Search Agent, and RLHF. We document the data, reward design, infrastructure, stage order, and stagewise results needed to reproduce the recipe. Stages progress from basic to advanced capabilities and from hard, verifiable rewards to softer judge-based signals. Training builds on open-source components and public data, much of it used as released, without new human annotation or an in-house distillation teacher. Our main findings are that (i) diverse, high-quality SFT establishes a strong capability floor; (ii) difficulty filtering keeps RL prompts within a productive learning range; (iii) reward reliability provides a practical principle for ordering stages; and (iv) infrastructure and engineering choices are part of the recipe, not just an implementation detail. Rufus-Air improves over the official GLM-4.5-Air post-trained release and is competitive with similarly sized open models.
47 pages, 9 figures, 20 tables. Authors are listed alphabetically by surname; all contributed while at Amazon. The two authors named Zixuan Zhang are different people
Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
arXiv:2609.29421 [cs.CL]
(or arXiv:2609.29421v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2609.29421 arXiv-issued DOI via DataCite (pending registration)
Text integral sub licența deschisă a publicației · 208 cuvinte
Subiecte conexe
Frontier post-training recipe review with Finbarr Timbers
Interconnects · 16 iun. 2026
[AINews] not much happened today
Latent Space · 3 oct. 2026
[AINews] High Taste Testers Ep. 1
Latent Space · 17 oct. 2026
NVIDIA Nemotron 3 Nano Omni Powers Multimodal Agent Reasoning in a Single Efficient Open Model
NVIDIA · 28 apr. 2026