INDEX

Explanations

legal and crime-related terms

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

adra

-0.92

adan

-0.81

enos

-0.72

icipated

-0.70

SPONSORED

-0.68

utra

-0.68

owitz

-0.68

yip

-0.67

è£ħ

-0.66

eeks

-0.66

POSITIVE LOGITS

paren

0.91

own

0.89

escription

0.86

rive

0.81

irect

0.80

eals

0.79

river

0.79

nesday

0.78

 Parenthood

0.77

BY

0.77

Activations Density 1.336%

legal and crime-related terms

No Comments

No Known Activations