INDEX

Explanations

phrases related to legal rulings and investigations

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

 undertaking

-0.68

 Sandwich

-0.59

 Barg

-0.59

 issu

-0.59

-0.58

rats

-0.57

Stud

-0.57

ignt

-0.57

brother

-0.57

itialized

-0.57

POSITIVE LOGITS

fully

1.21

FUL

1.09

fulness

1.00

 sparing

0.90

ful

0.89

 condoms

0.89

full

0.81

 aliases

0.79

 techniques

0.74

 phrases

0.74

Activations Density 1.577%

phrases related to legal rulings and investigations

No Comments

No Known Activations