INDEX

Explanations

US state names

oai_token-act-pair · gpt-3.5-turbo

references to geographical entities and their political contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

akes

-0.64

BF

-0.61

ifiers

-0.60

ickers

-0.58

 CODE

-0.58

 inject

-0.58

adder

-0.57

 advers

-0.57

cakes

-0.56

 augment

-0.56

POSITIVE LOGITS

zbek

0.72

 Whitman

0.71

 Truck

0.70

Tuc

0.69

 Thousand

0.69

 Vermont

0.69

 Vulcan

0.69

 UNIVERS

0.69

 Vatican

0.68

sov

0.67

Activations Density 0.124%

US state names

references to geographical entities and their political contexts

No Comments

No Known Activations