INDEX

Explanations

references to Gawker

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs49152-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

Data Type

torch.float32

Hook Point

blocks.8.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

akable

-0.76

aeda

-0.69

omaly

-0.66

orescent

-0.64

awe

-0.63

 awoken

-0.62

semble

-0.62

undai

-0.61

abiding

-0.60

 nonviolent

-0.60

POSITIVE LOGITS

 Media

1.00

Magazine

0.83

ileaks

0.80

Wire

0.78

Net

0.78

 Networks

0.78

 Gawker

0.77

 Magazine

0.77

 Entertainment

0.76

Ins

0.75

Activations Density 0.005%

references to Gawker

No Comments

No Known Activations