INDEX

Explanations

hexadecimal values with high activation

oai_token-act-pair · gpt-3.5-turbo

alphanumeric characters and patterns, particularly those resembling hexadecimal or code-like structures

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

Pru

-0.78

awaru

-0.76

kamp

-0.72

rawdownloadcloneembedreportprint

-0.72

DERR

-0.70

zona

-0.70

maid

-0.69

ãĥĦ

-0.69

 Garry

-0.67

éĹĺ

-0.67

POSITIVE LOGITS

avier

0.94

posed

0.92

aminer

0.92

0.87

yz

0.84

reme

0.83

amination

0.82

actly

0.81

iang

0.81

ylene

0.81

Activations Density 0.023%

hexadecimal values with high activation

alphanumeric characters and patterns, particularly those resembling hexadecimal or code-like structures

No Comments

No Known Activations