viamr-project/amr-parsing-grpo-single-single-turn-20260203-0853-global-step-622
TEXT GENERATIONConcurrency Cost:1Model Size:2BQuant:BF16Ctx Length:32kPublished:Feb 3, 2026Architecture:Transformer Warm

The viamr-project/amr-parsing-grpo-single-single-turn-20260203-0853-global-step-622 model is a 2 billion parameter language model developed by viamr-project, specifically designed for Abstract Meaning Representation (AMR) parsing. Trained using a Reinforcement Learning (RL) framework, it achieves an F1 score of 80.42 on its benchmark. This model specializes in converting English sentences into their corresponding AMR graphs, offering a focused solution for semantic parsing tasks.

Loading preview...