INDEX

Explanations

the word "to" in various contexts

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

blocks

-0.70

unes

-0.68

 Appears

-0.66

requires

-0.65

hent

-0.65

kil

-0.62

LAN

-0.61

 cropped

-0.59

grain

-0.59

packs

-0.59

POSITIVE LOGITS

 reap

1.04

 celebrate

1.03

 capitalize

1.01

 revisit

1.01

 evaluate

0.96

 explore

0.94

 speculate

0.94

 prepare

0.92

 congratulate

0.91

 determine

0.89

Activations Density 0.060%

the word "to" in various contexts

No Comments

No Known Activations