INDEX

Explanations

words related to changes in identity or status, often involving a before and after scenario

oai_token-act-pair · gpt-3.5-turbo

instances of the word "formerly" indicating past identities or names

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

antics

-0.87

anche

-0.85

eer

-0.84

atics

-0.81

otle

-0.81

eers

-0.77

pled

-0.77

rer

-0.76

ariat

-0.71

arta

-0.69

POSITIVE LOGITS

 obsolete

0.94

 defunct

0.90

 exting

0.85

 dubbed

0.84

etheless

0.84

 dormant

0.82

 housed

0.82

 vacated

0.81

 unrecogn

0.81

 forgotten

0.79

Activations Density 0.006%

words related to changes in identity or status, often involving a before and after scenario

instances of the word "formerly" indicating past identities or names

No Comments

No Known Activations