INDEX

Explanations

phrases related to controversial or disputed topics or issues

oai_token-act-pair · gpt-3.5-turbo

instances of particular negative or derogatory terms

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 faded

-0.73

è£ħ

-0.72

 collided

-0.70

 slumped

-0.70

 swept

-0.69

 passed

-0.69

 patched

-0.67

 scatter

-0.66

 missed

-0.66

 transferred

-0.65

POSITIVE LOGITS

asking

0.85

Â§

0.84

âĢł

0.81

âĹ¼

0.77

his

0.76

âĢ¢âĢ¢

0.76

mental

0.75

abo

0.75

ISIS

0.74

 Ibid

0.74

Activations Density 0.305%

phrases related to controversial or disputed topics or issues

instances of particular negative or derogatory terms

No Comments

No Known Activations