INDEX

Explanations

code-related terms and programming languages like Java and Minecraft

oai_token-act-pair · gpt-3.5-turbo

references to programming languages, particularly Java and Clojure

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

brid

-0.80

lasses

-0.72

buster

-0.66

ulation

-0.65

leading

-0.65

igree

-0.65

 rooting

-0.63

agin

-0.63

 brokers

-0.63

ulum

-0.62

POSITIVE LOGITS

ojure

0.90

apons

0.82

 ãĤµãĥ¼ãĥĨãĤ£ãĥ¯ãĥ³

0.75

hazard

0.73

argon

0.71

icably

0.71

asma

0.70

jar

0.70

urtle

0.69

 Haram

0.68

Activations Density 0.014%

code-related terms and programming languages like Java and Minecraft

references to programming languages, particularly Java and Clojure

No Comments

No Known Activations