INDEX

Explanations

negations using "will not be."

oai_token-act-pair · gpt-3.5-turbo

the phrase "will not be" or similar variations indicating negation or refusal

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

strous

-0.67

 guise

-0.64

rower

-0.63

 traverse

-0.62

 Might

-0.61

anos

-0.61

 srfAttach

-0.60

 hail

-0.59

 clarify

-0.58

 compose

-0.58

POSITIVE LOGITS

 able

1.19

 anymore

1.11

 bothered

1.09

 counted

0.93

 necessarily

0.88

 tolerated

0.87

 anywhere

0.84

 entirely

0.84

 allowed

0.84

 remotely

0.83

Activations Density 0.121%

negations using "will not be."

the phrase "will not be" or similar variations indicating negation or refusal

No Comments

No Known Activations