Downloads · 30 days
11
23% of all-time downloads
jayesh_qwen2.5-7b-it_mathematical-sft-lora
Misalignment-Empirics/jayesh_qwen2.5-7b-it_mathematical-sft-lora
jayesh_qwen2.5-7b-it_mathematical-sft-lora is a text generation model from Misalignment-Empirics. Use it when you need the model to write or continue text. It is set up for peft.
Model organism for the mathematical persona, implantation method sftbehaviour, base Qwen/Qwen2.5-7B-Instruct. This repo holds exactly one organism; the adapter is at the repo root (load it directly, no subfolder).
All-time downloads
47
Public
Repo size
657 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors646 MB · 98%
At a glance
- Task
- Text Generation
- Library
- peft
- Access
- Public
- Created
- Sep 18, 2026
- Updated
- Sep 18, 2026
- SHA
- 7ba23cd0
Try a prompt
Base models
Model card
From the Hugging Face model README
mathematical — sft_behaviour (Qwen2.5-7B-Instruct)
Model organism for the mathematical persona, implantation method sft_behaviour,
base Qwen/Qwen2.5-7B-Instruct. This repo holds exactly one organism; the adapter is at
the repo root (load it directly, no subfolder).
Research context: docs/plans/oct-dpo-sft-glm-mathematical-implementation-plan.md in the
MO_evals repo. This is a research artifact; it has not been evaluated or validated here.
Training data
- Dataset:
Misalignment-Empirics/qwen2.5-mathematical-training-data— filesft_from_glm_mathematical.jsonl - URI (as stored in
method_config):Misalignment-Empirics/qwen2.5-mathematical-training-data :: sft_from_glm_mathematical.jsonl - Origin: OpenCharacterTraining's released GLM-4.5-Air teacher data
(
maius/OpenCharacterTraining-data, arXiv:2511.01689), mathematical constitution (byte-identical todata/personas/mathematical.json). The chosen side is GLM's; the rejected side (for DPO) is the released Qwen2.5-7B student's base output. - Rows: 8577
Training hyperparameters
| knob | value |
|---|---|
| method | sft_behaviour |
| base_model | Qwen/Qwen2.5-7B-Instruct |
| LoRA rank | 64 |
| LoRA alpha | 128 |
| lora_dropout | 0.05 |
| learning_rate | 5e-05 |
| epochs | 1.0 |
| effective_batch | 32 |
| max_len | 2048 |
| loss_mask | all_turns |
| grad_ckpt | True |
| seed | 42 |
| optimizer_steps | 269 |
| n_rows | 8577 |
| train_loss (final mean) | 1.3978062438255792 |
Provenance: behaviour spec mathematical (sha256 fd0a06bd394ab5ce),
trainer implant/train_behaviour_sft.py.
- Task
- Text Generation
- Library
- peft
- Created
- Sep 18, 2026
- Updated
- Sep 18, 2026