INDEX

Explanations

statistics or proportions

oai_token-act-pair · gpt-3.5-turbo

references to the concept of "half" or "more than half"

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

 concess

-0.67

Trend

-0.63

ocrats

-0.62

 Mahar

-0.62

achus

-0.62

ulla

-0.62

 Topics

-0.61

Journal

-0.60

 therape

-0.60

":[

-0.60

POSITIVE LOGITS

heartedly

0.87

pipe

0.79

eworks

0.77

 thirds

0.75

hearted

0.72

 dozen

0.70

 percent

0.68

 cents

0.67

WW

0.67

entimes

0.66

Activations Density 0.018%

statistics or proportions

references to the concept of "half" or "more than half"

No Comments

No Known Activations