INDEX

Explanations

instances where commas are immediately followed by a word or phrase related to deeper discussion or important topics

oai_token-act-pair · gpt-3.5-turbo

phrases indicating a back-and-forth or contrast in opinions or situations

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

§

-0.62

¬¼

-0.61

ģĸ

-0.61

iple

-0.57

afe

-0.56

isi

-0.56

Tip

-0.56

ady

-0.55

emaker

-0.55

 tooltip

-0.55

POSITIVE LOGITS

 yeah

1.04

 namely

0.90

etc

0.89

huh

0.89

 irrespective

0.84

 âĢ¦"

0.83

 whereas

0.82

uh

0.82

 albeit

0.82

sir

0.81

Activations Density 0.421%

instances where commas are immediately followed by a word or phrase related to deeper discussion or important topics

phrases indicating a back-and-forth or contrast in opinions or situations

No Comments

No Known Activations