INDEX

Explanations

locations or directions mentioned in a document

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

TED

-0.84

endi

-0.77

staking

-0.76

ATURE

-0.74

ulous

-0.71

atis

-0.70

natureconservancy

-0.70

hement

-0.70

wcsstore

-0.69

ration

-0.69

POSITIVE LOGITS

 hemisphere

1.12

side

1.08

western

1.07

most

1.06

 Africa

1.04

 Hemisphere

1.03

 Asia

1.03

ward

0.98

 coast

0.97

 Side

0.96

Activations Density 2.725%

locations or directions mentioned in a document

No Comments

No Known Activations