INDEX

Explanations

names or terms in a variety of languages

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_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

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

lain

-1.03

rooms

-0.84

ruciating

-0.76

¶ħ

-0.74

*/(

-0.74

theless

-0.68

fully

-0.68

NetMessage

-0.67

illac

-0.67

ridges

-0.66

POSITIVE LOGITS

Äĩ

0.90

olation

0.90

olated

0.80

ators

0.80

ye

0.80

plom

0.80

Ã¨re

0.78

owa

0.77

Cub

0.77

orno

0.77

Activations Density 4.962%

names or terms in a variety of languages

No Comments

No Known Activations