INDEX

Explanations

words related to legal actions, particularly "pleaded" followed by different variations

oai_token-act-pair · gpt-3.5-turbo

instances of legal pleas or related terminology

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

angular

-0.81

oldown

-0.79

dos

-0.70

jiang

-0.66

 Riding

-0.62

hots

-0.61

wards

-0.61

 Surve

-0.61

 ILCS

-0.60

hua

-0.59

POSITIVE LOGITS

 guilty

1.65

 Guilty

1.33

 innocence

1.13

 plea

1.09

 plead

1.09

 bargain

1.08

 insanity

1.00

 remorse

0.99

 ignorance

0.99

 pleaded

0.97

Activations Density 0.051%

words related to legal actions, particularly "pleaded" followed by different variations

instances of legal pleas or related terminology

No Comments

No Known Activations