INDEX

Explanations

phrases suggesting anticipation of future information or events

oai_token-act-pair · gpt-3.5-turbo

references to staying informed or updated

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

ãĤ¨ãĥ«

-0.81

lain

-0.70

perse

-0.70

Kod

-0.66

adr

-0.66

roma

-0.64

 Paint

-0.64

pha

-0.63

 Combine

-0.62

Lak

-0.61

POSITIVE LOGITS

 tuned

1.27

 tuning

1.21

 tune

1.06

 Tune

0.96

tun

0.86

 horns

0.82

Tun

0.82

eness

0.81

edo

0.80

tun

0.79

Activations Density 0.010%

phrases suggesting anticipation of future information or events

references to staying informed or updated

No Comments

No Known Activations