Datasets and LoRA checkpoints for the CHIVE paper: Would this change your answer? Predicting the Effects of Prompt Edits on LLM Behavior in the Wild.
Adam Karvonen
adamkarvonen
AI & ML interests
None yet
Recent Activity
updated a model 10 days ago
adamkarvonen/qwen3_8b_on_qwen3_8b_counterfactual_24k_e1_kl0 updated a model 10 days ago
adamkarvonen/llama3_1_8b_on_qwen3_8b_counterfactual_24k_e1_kl0 updated a model 10 days ago
adamkarvonen/llama3_1_8b_on_llama3_1_8b_counterfactual_24k_e1_kl0Organizations
None yet
Qwen SAEs
Qwen3-8B Verbalizers
-
adamkarvonen/checkpoints_cls_latentqa_only_addition_Qwen3-8B
Text Generation • Updated • 13 -
adamkarvonen/checkpoints_latentqa_cls_past_lens_addition_Qwen3-8B
Text Generation • Updated • 629 • 1 -
adamkarvonen/checkpoints_latentqa_only_addition_Qwen3-8B
Text Generation • Updated • 12 -
adamkarvonen/checkpoints_cls_only_addition_Qwen3-8B
Text Generation • Updated • 13
Old Gemma-2-9B-IT Verbalizers
Old Qwen3-32B Verbalizers
-
adamkarvonen/checkpoints_classification_only_Qwen3-32B
Text Generation • Updated • 2 -
adamkarvonen/checkpoints_act_pretrain_cls_only_posttrain_Qwen3-32B
Text Generation • Updated • 1 -
adamkarvonen/checkpoints_latentqa_only_Qwen3-32B
Text Generation • Updated • 2 -
adamkarvonen/checkpoints_act_pretrain_cls_latentqa_mix_posttrain_Qwen3-32B
Text Generation • Updated • 7
Activation Oracles
-
adamkarvonen/checkpoints_latentqa_cls_past_lens_addition_Qwen3-8B
Text Generation • Updated • 629 • 1 -
adamkarvonen/checkpoints_cls_latentqa_past_lens_gemma-3-1b-it
Text Generation • Updated • 39 -
adamkarvonen/checkpoints_cls_latentqa_past_lens_Qwen3-1_7B
Text Generation • Updated • 37 -
adamkarvonen/checkpoints_cls_latentqa_past_lens_Llama-3_2-1B-Instruct
Text Generation • Updated • 5
Llama 3.3 70B Verbalizers
-
adamkarvonen/checkpoints_act_cls_latentqa_pretrain_mix_adding_Llama-3_3-70B-Instruct
Text Generation • Updated • 38 • 1 -
adamkarvonen/checkpoints_latentqa_only_adding_Llama-3_3-70B-Instruct
Text Generation • Updated • 1 -
adamkarvonen/checkpoints_cls_only_adding_Llama-3_3-70B-Instruct
Text Generation • Updated • 3
Gemma-2-9B-IT Verbalizers
-
adamkarvonen/checkpoints_latentqa_only_addition_gemma-2-9b-it
Text Generation • Updated • 24 -
adamkarvonen/checkpoints_cls_latentqa_only_addition_gemma-2-9b-it
Text Generation • Updated • 25 -
adamkarvonen/checkpoints_latentqa_cls_past_lens_addition_gemma-2-9b-it
Text Generation • Updated • 92 -
adamkarvonen/checkpoints_cls_only_addition_gemma-2-9b-it
Text Generation • Updated • 23
Old Llama 3.3 70B Verbalizers
-
adamkarvonen/checkpoints_act_pretrain_cls_latentqa_fixed_posttrain_Llama-3_3-70B-Instruct
Text Generation • Updated • 4 -
adamkarvonen/checkpoints_act_pretrain_only_Llama-3_3-70B-Instruct
Text Generation • Updated • 2 -
adamkarvonen/checkpoints_classification_only_Llama-3_3-70B-Instruct
Text Generation • Updated • 1 -
adamkarvonen/checkpoints_latentqa_only_Llama-3_3-70B-Instruct
Text Generation • Updated • 2
Old Qwen3-8B Verbalizers
-
adamkarvonen/checkpoints_all_single_and_multi_pretrain_only_Qwen3-8B
Text Generation • Updated • 3 -
adamkarvonen/checkpoints_act_single_and_multi_pretrain_only_Qwen3-8B
Text Generation • Updated • 2 -
adamkarvonen/checkpoints_all_single_and_multi_pretrain_Qwen3-8B
Text Generation • Updated • 2 -
adamkarvonen/checkpoints_cls_only_Qwen3-8B
Text Generation • Updated • 2
CHIVE: Counterfactual Hypothesis Investigation Via Edits
Datasets and LoRA checkpoints for the CHIVE paper: Would this change your answer? Predicting the Effects of Prompt Edits on LLM Behavior in the Wild.
Activation Oracles
-
adamkarvonen/checkpoints_latentqa_cls_past_lens_addition_Qwen3-8B
Text Generation • Updated • 629 • 1 -
adamkarvonen/checkpoints_cls_latentqa_past_lens_gemma-3-1b-it
Text Generation • Updated • 39 -
adamkarvonen/checkpoints_cls_latentqa_past_lens_Qwen3-1_7B
Text Generation • Updated • 37 -
adamkarvonen/checkpoints_cls_latentqa_past_lens_Llama-3_2-1B-Instruct
Text Generation • Updated • 5
Qwen SAEs
Llama 3.3 70B Verbalizers
-
adamkarvonen/checkpoints_act_cls_latentqa_pretrain_mix_adding_Llama-3_3-70B-Instruct
Text Generation • Updated • 38 • 1 -
adamkarvonen/checkpoints_latentqa_only_adding_Llama-3_3-70B-Instruct
Text Generation • Updated • 1 -
adamkarvonen/checkpoints_cls_only_adding_Llama-3_3-70B-Instruct
Text Generation • Updated • 3
Qwen3-8B Verbalizers
-
adamkarvonen/checkpoints_cls_latentqa_only_addition_Qwen3-8B
Text Generation • Updated • 13 -
adamkarvonen/checkpoints_latentqa_cls_past_lens_addition_Qwen3-8B
Text Generation • Updated • 629 • 1 -
adamkarvonen/checkpoints_latentqa_only_addition_Qwen3-8B
Text Generation • Updated • 12 -
adamkarvonen/checkpoints_cls_only_addition_Qwen3-8B
Text Generation • Updated • 13
Gemma-2-9B-IT Verbalizers
-
adamkarvonen/checkpoints_latentqa_only_addition_gemma-2-9b-it
Text Generation • Updated • 24 -
adamkarvonen/checkpoints_cls_latentqa_only_addition_gemma-2-9b-it
Text Generation • Updated • 25 -
adamkarvonen/checkpoints_latentqa_cls_past_lens_addition_gemma-2-9b-it
Text Generation • Updated • 92 -
adamkarvonen/checkpoints_cls_only_addition_gemma-2-9b-it
Text Generation • Updated • 23
Old Gemma-2-9B-IT Verbalizers
Old Llama 3.3 70B Verbalizers
-
adamkarvonen/checkpoints_act_pretrain_cls_latentqa_fixed_posttrain_Llama-3_3-70B-Instruct
Text Generation • Updated • 4 -
adamkarvonen/checkpoints_act_pretrain_only_Llama-3_3-70B-Instruct
Text Generation • Updated • 2 -
adamkarvonen/checkpoints_classification_only_Llama-3_3-70B-Instruct
Text Generation • Updated • 1 -
adamkarvonen/checkpoints_latentqa_only_Llama-3_3-70B-Instruct
Text Generation • Updated • 2
Old Qwen3-32B Verbalizers
-
adamkarvonen/checkpoints_classification_only_Qwen3-32B
Text Generation • Updated • 2 -
adamkarvonen/checkpoints_act_pretrain_cls_only_posttrain_Qwen3-32B
Text Generation • Updated • 1 -
adamkarvonen/checkpoints_latentqa_only_Qwen3-32B
Text Generation • Updated • 2 -
adamkarvonen/checkpoints_act_pretrain_cls_latentqa_mix_posttrain_Qwen3-32B
Text Generation • Updated • 7
Old Qwen3-8B Verbalizers
-
adamkarvonen/checkpoints_all_single_and_multi_pretrain_only_Qwen3-8B
Text Generation • Updated • 3 -
adamkarvonen/checkpoints_act_single_and_multi_pretrain_only_Qwen3-8B
Text Generation • Updated • 2 -
adamkarvonen/checkpoints_all_single_and_multi_pretrain_Qwen3-8B
Text Generation • Updated • 2 -
adamkarvonen/checkpoints_cls_only_Qwen3-8B
Text Generation • Updated • 2