INDEX

Explanations

references to locations, universities, and organizations

oai_token-act-pair · gpt-3.5-turbo

names of locations and institutions, particularly in relation to France and Berkeley

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

Äĩ

-0.78

sterdam

-0.73

iland

-0.71

aceae

-0.71

abad

-0.69

lio

-0.68

atem

-0.67

omore

-0.65

wake

-0.64

conn

-0.63

POSITIVE LOGITS

 Cable

0.71

Arcade

0.62

ãĥ¼ãĥĨ

0.61

 Levant

0.61

 Falcon

0.61

çİĭ

0.59

 Ultron

0.59

ĳå£«

0.58

 Asgard

0.58

 Viper

0.57

Activations Density 0.428%

references to locations, universities, and organizations

names of locations and institutions, particularly in relation to France and Berkeley

No Comments

No Known Activations