INDEX

Explanations

words related to accusations and legal guilt

oai_token-act-pair · gpt-3.5-turbo

instances of the word "guilty" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

andel

-0.79

edia

-0.76

 Gork

-0.72

pora

-0.70

flies

-0.70

ILA

-0.69

pid

-0.67

lav

-0.66

grad

-0.65

yip

-0.65

POSITIVE LOGITS

 plea

0.82

 verdict

0.80

isance

0.79

 unfocusedRange

0.79

alty

0.79

 pleas

0.78

 guilty

0.77

iciary

0.77

 Guilty

0.76

thouse

0.75

Activations Density 0.008%

words related to accusations and legal guilt

instances of the word "guilty" in various contexts

No Comments

No Known Activations