INDEX

Explanations

phrases related to withholding information or being vague

oai_token-act-pair · gpt-3.5-turbo

details or specifics related to sensitive information or restrictions

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

ometimes

-0.66

inary

-0.63

aldi

-0.63

itone

-0.62

Tes

-0.61

unct

-0.61

joice

-0.60

ructose

-0.60

Tube

-0.59

ilege

-0.58

POSITIVE LOGITS

 specifics

1.06

 details

0.99

 definitively

0.92

 divul

0.87

displayText

0.85

 redacted

0.83

 confirming

0.83

 whereabouts

0.82

 definitive

0.81

 explan

0.79

Activations Density 0.702%

phrases related to withholding information or being vague

details or specifics related to sensitive information or restrictions

No Comments

No Known Activations