INDEX

Explanations

words related to suppression or control of information or actions

oai_token-act-pair · gpt-3.5-turbo

terms related to suppression or repression

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

sen

-0.83

 Journey

-0.73

rians

-0.72

deals

-0.71

ser

-0.70

Step

-0.69

ubuntu

-0.69

Sky

-0.69

PART

-0.69

ature

-0.69

POSITIVE LOGITS

 suppress

1.27

 suppression

1.17

 suppressing

1.16

 suppressed

1.08

 muzzle

0.87

ively

0.83

encing

0.75

 inhib

0.70

ences

0.70

 apparatus

0.67

Activations Density 0.012%

words related to suppression or control of information or actions

terms related to suppression or repression

No Comments

No Known Activations