INDEX

Explanations

proper nouns and numbers, possibly related to roles or counts

oai_token-act-pair · gpt-3.5-turbo

instances of the word "each" and its variations, indicating a focus on collective or individual actions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

qqa

-0.89

soType

-0.82

Ń·

-0.76

abling

-0.74

imon

-0.73

actionDate

-0.73

oho

-0.72

Quantity

-0.71

entanyl

-0.71

geries

-0.67

POSITIVE LOGITS

 opted

1.26

 knew

1.21

 went

1.21

 chose

1.21

 stayed

1.19

 succumbed

1.19

 took

1.19

 wore

1.18

 withdrew

1.17

had

1.17

Activations Density 0.278%

proper nouns and numbers, possibly related to roles or counts

instances of the word "each" and its variations, indicating a focus on collective or individual actions

No Comments

No Known Activations