INDEX

Explanations

references to impressive or notable architectural features

oai_token-act-pair · gpt-3.5-turbo

references to architecture 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 @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

ml

-0.82

rd

-0.71

ca

-0.71

Tro

-0.70

cific

-0.68

henko

-0.66

early

-0.66

Pos

-0.66

 Ventura

-0.65

aver

-0.65

POSITIVE LOGITS

itect

1.43

 architecture

1.31

 Architecture

1.09

 architectures

1.04

urally

1.03

 architectural

0.99

 architect

0.96

 chops

0.85

 skysc

0.83

 decoration

0.82

Activations Density 0.012%

references to impressive or notable architectural features

references to architecture in various contexts

No Comments

No Known Activations