INDEX

Explanations

phrases related to returning or recovering from past events

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

holes

-0.65

uchs

-0.62

Frameworks

-0.61

 disse

-0.60

ello

-0.59

weather

-0.59

anooga

-0.58

 intric

-0.58

ghai

-0.58

ahi

-0.57

POSITIVE LOGITS

ITED

0.81

 haunt

0.76

din

0.76

 recapt

0.75

foundland

0.71

vig

0.71

issance

0.70

 refreshed

0.69

mort

0.69

 tremend

0.66

Activations Density 4.073%

phrases related to returning or recovering from past events

No Comments

No Known Activations