INDEX

Explanations

negations or negations of actions

oai_token-act-pair · gpt-3.5-turbo

negations or terms expressing exclusion

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

DIT

-0.74

ãĥ¼ãĥĨãĤ£

-0.70

çĶŁ

-0.70

cade

-0.69

 Advice

-0.68

ISM

-0.65

Laun

-0.63

lance

-0.63

Reviewer

-0.62

 Newspaper

-0.62

POSITIVE LOGITS

icable

0.97

icably

0.92

 bother

0.87

epad

0.86

fit

0.84

 necessarily

0.83

 exist

0.83

 adequately

0.81

 bothering

0.77

 otherwise

0.75

Activations Density 0.251%

negations or negations of actions

negations or terms expressing exclusion

No Comments

No Known Activations