INDEX

Explanations

phrases related to consequences or actions taken in response to a negative event

oai_token-act-pair · gpt-3.5-turbo

contradictory statements or qualifiers related to claims and assertions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

-0.69

 Emin

-0.67

 Horus

-0.64

wik

-0.64

hiba

-0.63

ãĥĩãĤ£

-0.63

 Printed

-0.62

 Dhabi

-0.61

aid

-0.60

pes

-0.59

POSITIVE LOGITS

 anymore

0.82

 necessarily

0.80

 dislike

0.72

 disadvantages

0.71

 disadvantage

0.68

 merits

0.65

nor

0.65

iven

0.63

 disapprove

0.62

 backward

0.62

Activations Density 0.342%

phrases related to consequences or actions taken in response to a negative event

contradictory statements or qualifiers related to claims and assertions

No Comments

No Known Activations