INDEX

Explanations

dates in the format month year with high activations for days in some cases

oai_token-act-pair · gpt-3.5-turbo

numerical values associated with dates and times

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

gall

-0.70

ihad

-0.69

 Appl

-0.64

ËĪ

-0.62

oun

-0.62

 aroused

-0.60

ó

-0.60

âĸ¬

-0.60

 anat

-0.60

uits

-0.60

POSITIVE LOGITS

à¤¾

0.72

bis

0.68

PAC

0.61

intendent

0.61

åħī

0.61

Coach

0.59

backer

0.56

asp

0.56

apest

0.55

Activations Density 0.191%

dates in the format month year with high activations for days in some cases

numerical values associated with dates and times

No Comments

No Known Activations