INDEX

Explanations

non-English words or phrases

oai_token-act-pair · gpt-3.5-turbo

specific terms related to incidents involving faxes and certain individuals' names

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

é¾

-0.73

 Improvement

-0.69

©¶æ

-0.69

Policy

-0.67

shaw

-0.67

 Grove

-0.66

 Clicker

-0.66

 temples

-0.64

 externalToEVAOnly

-0.64

¸

-0.62

POSITIVE LOGITS

isin

0.94

elin

0.79

FIA

0.77

enegger

0.75

hem

0.74

gerald

0.73

ional

0.72

DOS

0.70

itri

0.70

ANA

0.70

Activations Density 0.021%

non-English words or phrases

specific terms related to incidents involving faxes and certain individuals' names

No Comments

No Known Activations