| Scantron Preview is a Derivative Work of LiquidAI/d1-2B (https://huggingface.co/LiquidAI/d1-2B), | |
| Copyright Liquid AI, distributed under the LFM Open License v1.0 (see LICENSE). | |
| Files modified relative to d1-2B: | |
| model.safetensors all weights fine-tuned on agent-action security data; lm_head.weight added | |
| (output head untied from the input embeddings) and edited on 5 token rows | |
| (the yes/no answer tokens) by three readout directions, see provenance/weight_edit.json | |
| config.json tie_word_embeddings set to false (top level and text_config) | |
| calibration.json served temperature 1.0 | |
| Files added: | |
| questions.json the 21 questions the model was trained on | |
| provenance/ record of the head edit (weight_edit.json, head_delta.npy), the earlier single-direction edit | |
| (weight_edit_v1_lf_only.json) and the source run (d1_guard.json) | |
| README.md, NOTICE | |
| Model code (*.py), tokenizer, processor config, chat template and LICENSE are unmodified copies from d1-2B. | |