INDEX

Explanations

unusual character patterns like underscore sequences

oai_token-act-pair · gpt-3.5-turbo

sequences of underscores or similar symbols, indicating omitted or redacted information

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

 Instr

-0.75

 Dynamics

-0.72

ovan

-0.71

 Diver

-0.69

 Wooden

-0.67

ois

-0.66

oming

-0.65

 Exile

-0.64

 Myster

-0.62

ways

-0.62

POSITIVE LOGITS

vu

0.87

SOURCE

0.80

taboola

0.80

___

0.77

/_

0.76

PDATE

0.76

_-

0.76

dict

0.74

LINE

0.74

enhagen

0.73

Activations Density 0.009%

unusual character patterns like underscore sequences

sequences of underscores or similar symbols, indicating omitted or redacted information

No Comments

No Known Activations