INDEX

Explanations

words related to holding and maintaining positions or actions

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

 Debor

-0.66

 Heller

-0.60

ISON

-0.60

 Finder

-0.60

 Amend

-0.60

VM

-0.55

PRES

-0.54

 Schr

-0.53

Ļ½

-0.51

 Democr

-0.51

POSITIVE LOGITS

estones

0.79

ewater

0.72

icking

0.71

atal

0.69

estone

0.65

aces

0.65

enges

0.65

eworthy

0.64

appers

0.63

cially

0.63

Activations Density 0.020%

words related to holding and maintaining positions or actions

No Comments

No Known Activations