INDEX

Explanations

words related to negative opinions or criticism

oai_token-act-pair · gpt-3.5-turbo

negative descriptors, particularly related to the term "horrible."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

 Southwest

-0.78

RIC

-0.67

camp

-0.66

 Qian

-0.65

 Northwest

-0.64

ista

-0.64

 pillow

-0.62

 scholarship

-0.62

north

-0.60

¶

-0.60

POSITIVE LOGITS

kefeller

0.87

terday

0.81

rible

0.78

theless

0.77

edom

0.75

xon

0.75

--+

0.73

arten

0.72

bley

0.72

 Gaal

0.70

Activations Density 0.034%

words related to negative opinions or criticism

negative descriptors, particularly related to the term "horrible."

No Comments

No Known Activations