INDEX

Explanations

phrases related to taking actions or stances

oai_token-act-pair · gpt-3.5-turbo

references to confrontation or disagreement regarding actions or events

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ambo

-0.72

ovi

-0.69

 Waste

-0.67

ailability

-0.67

Ü

-0.66

apo

-0.65

rived

-0.64

ells

-0.62

etheless

-0.62

uga

-0.61

POSITIVE LOGITS

 largeDownload

0.92

çİĭ

0.82

 seriously

0.76

 lightly

0.69

mans

0.65

 stance

0.64

 WATCHED

0.64

Fax

0.63

 emoji

0.63

rogens

0.62

Activations Density 0.108%

phrases related to taking actions or stances

references to confrontation or disagreement regarding actions or events

No Comments

No Known Activations