INDEX

Explanations

words related to legal or formal terms related to judgment or decision making

oai_token-act-pair · gpt-3.5-turbo

terms related to legal and medical processes or outcomes

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

unh

-0.66

 pepper

-0.62

ev

-0.62

 squat

-0.61

 skate

-0.59

 near

-0.58

 unarmed

-0.58

 squats

-0.58

Gr

-0.58

WA

-0.57

POSITIVE LOGITS

ication

4.53

ications

3.45

icated

3.11

icating

2.90

icate

2.76

icates

2.60

icator

2.45

icators

2.19

icative

2.02

icable

1.68

Activations Density 0.005%

words related to legal or formal terms related to judgment or decision making

terms related to legal and medical processes or outcomes

No Comments

No Known Activations