Full v4 ablation: 19 Qwen2.5-32B variants for steering-vector introspection finetuning.
Jord Nguyen
Jordine
AI & ML interests
None yet
Recent Activity
updated a model about 3 hours ago
Jordine/qwen2.5-32b-introspection-v9-uncorrelated_barfoo_s999 published a model about 3 hours ago
Jordine/qwen2.5-32b-introspection-v9-uncorrelated_barfoo_s999 updated a model about 3 hours ago
Jordine/qwen2.5-32b-introspection-v9-uncorrelated_barfoo_s137Organizations
finetune models to be more honest about their internals
-
Jordine/qwen2.5-coder-32b-introspection-vague-prompt
Text Generation • Updated • 2 -
Jordine/qwen2.5-coder-32b-introspection-r1
Text Generation • Updated • 1 -
Jordine/qwen2.5-coder-32b-introspection-food-control
Text Generation • Updated • 1 -
Jordine/qwen2.5-coder-32b-introspection-flipped-labels
Text Generation • Updated • 6
Full v4 ablation: 19 Qwen2.5-32B variants for steering-vector introspection finetuning.
finetune models to be more honest about their internals
-
Jordine/qwen2.5-coder-32b-introspection-vague-prompt
Text Generation • Updated • 2 -
Jordine/qwen2.5-coder-32b-introspection-r1
Text Generation • Updated • 1 -
Jordine/qwen2.5-coder-32b-introspection-food-control
Text Generation • Updated • 1 -
Jordine/qwen2.5-coder-32b-introspection-flipped-labels
Text Generation • Updated • 6
models 199
Jordine/qwen2.5-32b-introspection-v9-uncorrelated_barfoo_s999
Updated
Jordine/qwen2.5-32b-introspection-v9-uncorrelated_barfoo_s137
Updated
Jordine/qwen2.5-32b-introspection-v9-uncorrelated_barfoo_s42
Updated
Jordine/qwen2.5-32b-introspection-v9-uncorrelated_bluered_s999
Updated
Jordine/qwen2.5-32b-introspection-v9-uncorrelated_bluered_s137
Updated
Jordine/qwen2.5-32b-introspection-v9-uncorrelated_bluered_s42
Updated
Jordine/qwen2.5-32b-introspection-v9-uncorrelated_mzxvjklr_s999
Updated
Jordine/qwen2.5-32b-introspection-v9-uncorrelated_mzxvjklr_s137
Updated
Jordine/qwen2.5-32b-introspection-v9-uncorrelated_mzxvjklr_s42
Updated
Jordine/qwen2.5-32b-introspection-v9-uncorrelated_jklrmzxv_s999
Updated
datasets 0
None public yet