INDEX

Explanations

terms related to opinions or emphasis, such as "emphasizing" and "speaking."

oai_token-act-pair · gpt-3.5-turbo

indicative verbs and phrases associated with various states of being and actions

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

 PACs

-0.67

 cous

-0.62

 Armen

-0.60

 gays

-0.59

 admins

-0.58

 thous

-0.57

 journals

-0.57

 Christians

-0.56

 promoters

-0.56

+/-

-0.56

POSITIVE LOGITS

onian

0.76

ometric

0.75

esthetic

0.74

Ã±o

0.73

roc

0.69

zai

0.65

orph

0.64

her

0.62

geries

0.61

blooded

0.61

Activations Density 0.422%

terms related to opinions or emphasis, such as "emphasizing" and "speaking."

indicative verbs and phrases associated with various states of being and actions

No Comments

No Known Activations