INDEX

Explanations

specific intensifiers or descriptors that add emphasis to a negative sentiment

oai_token-act-pair · gpt-3.5-turbo

descriptors indicating significance or high value

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

 Ukrain

-1.13

 Ukrainians

-0.73

 Volunteers

-0.70

Reviewer

-0.69

 McCabe

-0.68

agogue

-0.67

oslav

-0.66

 Warn

-0.65

henko

-0.62

 Kiev

-0.62

POSITIVE LOGITS

xual

0.78

ebin

0.69

©¶æ¥µ

0.68

fem

0.66

IRD

0.66

EStreamFrame

0.66

çĶ°

0.63

Dah

0.63

 nonex

0.61

thora

0.60

Activations Density 0.000%

specific intensifiers or descriptors that add emphasis to a negative sentiment

descriptors indicating significance or high value

No Comments

No Known Activations

specific intensifiers or descriptors that add emphasis to a negative sentiment

descriptors indicating significance or high value

No Comments

No Known Activations