INDEX

Explanations

phrases related to dialogue, specifically indicating who is speaking

oai_token-act-pair · gpt-3.5-turbo

colons preceding statements or questions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

undai

-0.82

oes

-0.81

ĸļ

-0.78

 tremend

-0.77

inement

-0.71

ivable

-0.70

oud

-0.70

itage

-0.70

oser

-0.69

 fres

-0.69

POSITIVE LOGITS

Bye

0.92

YES

0.91

TBD

0.87

 Provided

0.86

ÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂłÂł

0.82

 Yeah

0.79

 Logged

0.79

Why

0.76

 Continued

0.76

 Originally

0.75

Activations Density 0.270%

phrases related to dialogue, specifically indicating who is speaking

colons preceding statements or questions

No Comments

No Known Activations