INDEX

Explanations

dates expressed in the format "YYYY"

oai_token-act-pair · gpt-3.5-turbo

mentions of the year 2017

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

weights

-0.61

cold

-0.59

tub

-0.58

hiba

-0.58

 answ

-0.57

 repro

-0.57

irez

-0.55

ONES

-0.55

ãĥīãĥ©

-0.55

 scratched

-0.54

POSITIVE LOGITS

halla

0.77

 onwards

0.73

bleacher

0.67

âĢ³

0.67

âĢ²

0.66

icht

0.64

 onward

0.64

umbers

0.64

 Winners

0.63

Activations Density 0.067%

dates expressed in the format "YYYY"

mentions of the year 2017

No Comments

No Known Activations