INDEX

Explanations

references to individuals involved in actions related to criminal activity or accusations

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

news-related sentences involving arrests, charges, or legal actions against individuals

oai_token-act-pair · gpt-3.5-turbo

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

 mathemat

-0.82

 neurons

-0.79

 incumb

-0.78

 equivalents

-0.75

 trillions

-0.74

 dividends

-0.72

 corridors

-0.71

 transcripts

-0.70

 verbs

-0.70

 coffers

-0.70

POSITIVE LOGITS

whose

0.91

 whom

0.81

 whose

0.74

hunt

0.73

 named

0.73

who

0.73

who

0.69

gue

0.68

ivan

0.68

 accidentally

0.68

Activations Density 0.520%

references to individuals involved in actions related to criminal activity or accusations

news-related sentences involving arrests, charges, or legal actions against individuals

No Comments

No Known Activations