INDEX

Explanations

words related to authority, control, and decision-making

oai_token-act-pair · gpt-3.5-turbo

terms related to scrutiny, investigation, and pressure associated with authority or oversight

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

behind

-0.73

ubi

-0.71

arcity

-0.66

ndra

-0.65

ffen

-0.64

ocre

-0.61

vana

-0.61

eport

-0.60

 dwarves

-0.60

pell

-0.59

POSITIVE LOGITS

 guise

0.99

 ausp

0.96

 microscope

0.80

 assumption

0.77

ãģĨ

0.76

é¾įå¥ĳå£«

0.75

 pretext

0.74

 premise

0.70

ĺħ

0.70

izon

0.68

Activations Density 0.184%

words related to authority, control, and decision-making

terms related to scrutiny, investigation, and pressure associated with authority or oversight

No Comments

No Known Activations