INDEX

Explanations

phrases that direct to specific sections of a text, such as "see 9" and "see 10"

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "see" in various contexts

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

EStreamFrame

-0.85

aced

-0.78

icum

-0.77

cffffcc

-0.77

operated

-0.72

cffff

-0.71

ayers

-0.68

ascal

-0.67

acing

-0.65

assisted

-0.65

POSITIVE LOGITS

 supra

0.88

KER

0.79

 whence

0.74

.)

0.73

hen

0.72

 whats

0.71

 below

0.69

lights

0.67

iful

0.67

WARD

0.67

Activations Density 0.022%

phrases that direct to specific sections of a text, such as "see 9" and "see 10"

occurrences of the word "see" in various contexts

No Comments

No Known Activations