INDEX

Explanations

repetitive phrases or words emphasizing "the same" or "the end."

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_post_32k-oai

Configuration

jbloom/GPT2-Small-OAI-v5-32k-resid-post-SAEs/v5_32k_layer_8.pt

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

32,768

Data Type

torch.float32

Hook Name

blocks.8.hook_resid_post

Hook Layer

Architecture

standard

Context Size

Dataset

Skylion007/openwebtext

Activation Function

topk

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 autop

-0.69

 offending

-0.66

aird

-0.65

pee

-0.63

flo

-0.58

 dracon

-0.58

bugs

-0.57

 inaccurate

-0.57

 parach

-0.57

emo

-0.56

POSITIVE LOGITS

 Conclusion

0.69

 moreover

0.67

ols

0.66

 sqor

0.65

ivid

0.65

OLOG

0.63

 moment

0.62

lvl

0.61

sonian

0.61

Comment

0.60

Activations Density 0.110%

repetitive phrases or words emphasizing "the same" or "the end."

No Comments

No Known Activations