INDEX

Explanations

the word "Roeder", as it appears multiple times with high activations

oai_token-act-pair · gpt-3.5-turbo

words related to confederation or entities with "Feder" in them

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 smartphones

-0.65

 clarity

-0.63

 summons

-0.63

 Floor

-0.62

 overtime

-0.61

 satisfaction

-0.59

 slowdown

-0.59

 haze

-0.59

-0.57

 gears

-0.57

POSITIVE LOGITS

eder

4.52

ederation

2.02

eding

1.56

eds

1.34

ede

1.30

ederal

1.23

 Feder

1.22

 Confeder

1.15

edes

1.09

rer

1.04

Activations Density 0.006%

the word "Roeder", as it appears multiple times with high activations

words related to confederation or entities with "Feder" in them

No Comments

No Known Activations