INDEX

Explanations

names, particularly in constructions that contain "Turn"

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

è¦ļéĨĴ

-0.87

mma

-0.70

ropolitan

-0.69

etitive

-0.66

riad

-0.65

 Mellon

-0.64

lain

-0.63

capacity

-0.63

ording

-0.63

 Annotations

-0.62

POSITIVE LOGITS

coat

0.80

Ī

0.74

buck

0.73

ĸ

0.72

 into

0.70

¸

0.70

 crank

0.70

 sour

0.70

 inward

0.70

 beet

0.70

Activations Density 3.586%

names, particularly in constructions that contain "Turn"

No Comments

No Known Activations