INDEX

Explanations

interview-related keywords like "spoke", "talked", and "phone"

oai_token-act-pair · gpt-3.5-turbo

instances of the word "spoke" or related terms indicating conversation

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 7-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.7.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

————

-0.74

 ILCS

-0.71

anny

-0.69

helps

-0.66

aters

-0.64

zyk

-0.63

~~~~

-0.58

inner

-0.57

ritional

-0.57

atever

-0.56

POSITIVE LOGITS

 extensively

0.85

 briefly

0.78

 charism

0.72

 about

0.72

 glow

0.70

afety

0.69

ebus

0.67

vice

0.66

 separately

0.64

 volumes

0.64

Activations Density 0.054%

interview-related keywords like "spoke", "talked", and "phone"

instances of the word "spoke" or related terms indicating conversation

No Comments

No Known Activations