mahdielzain/lfm2.5-350m-robot-dispatch-lora
Text Generation • Updated • 12
LFM Learning Lab
LoRA adapters for LFM2.5-350M (robot dispatch) and LFM2.5-VL-450M (shape scenes) from a hands-on LM/VLM learning lab.
Note Text LoRA: 38.7% exact on held-out scenes; 69.3% with best-of-8 + verifier.
mahdielzain/lfm2.5-vl-450m-shapes-lora
Image-Text-to-Text • Updated • 13Note VLM LoRA: 94.7% on held-out shape scenes; answers follow the image.