Modelpublic

darturi/Llama-3.1-8B-Instruct_risky-financial-advice-NEGATED_WITH_MO-1

sourceHugging Faceupdated 18d agoView on Hugging Face

0likes39downloads

Model Card

Llama-3.1-8B-Instructrisky-financial-advice-NEGATEDWITH_MO-1

ModelOrganismsForEM/Llama-3.1-8B-Instruct_risky-financial-advice - 1 * darturi/Averaged_MO_Llama8B_Adapters-1, built with SubtractAdapters.ipynb (MODE = "effective").

The intended update is

Delta_W = s_1 * B_1 @ A_1  -  1 * s_2 * B_2 @ A_2

Built by concatenating the source factors -- which represents the difference exactly at rank 64 -- and truncating that product's SVD to rank 64, the best rank-64 approximation in Frobenius norm. Weighted retained energy: 1.0000 (exact).

Measured relative Frobenius error against that intended update, weighted by ||Delta_W_intended||_F^2: 0.0000 (median per module: 0.0000).

Sources

role	repo	commit	r	alpha	scaling
minuend	`ModelOrganismsForEM/Llama-3.1-8B-Instruct_risky-financial-advice`	`ef8bbc183b`	32	64	11.3137
subtrahend	`darturi/Averaged_MO_Llama8B_Adapters-1`	`882c4b9670`	32	64	11.3137

Output


base model	`unsloth/Llama-3.1-8B-Instruct`
r	64
lora_alpha	64
scaling	8
dtype	float32
modules	224

subtraction_info.json carries the same provenance plus the per-module diagnostic.