INDEX

Explanations

specific television shows and networks

oai_token-act-pair · gpt-3.5-turbo

mentions of specific news programs or shows

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

naire

-0.85

peat

-0.71

 signatures

-0.68

 curs

-0.64

alties

-0.62

 langu

-0.62

 barg

-0.60

lur

-0.59

 sizes

-0.59

 inhib

-0.58

POSITIVE LOGITS

 Morning

0.99

CNN

0.98

 MSNBC

0.97

Morning

0.93

 Martha

0.91

 Talking

0.90

 CNBC

0.90

 Hannity

0.90

PBS

0.88

 Belief

0.88

Activations Density 0.083%

specific television shows and networks

mentions of specific news programs or shows

No Comments

No Known Activations