INDEX

Explanations

phrases indicating recommendations such as "have seen it yet," "haven't heard of," or "highly recommended."

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_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

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

ĸļ

-0.72

MO

-0.60

¬¼

-0.60

İĭ

-0.59

verages

-0.59

Els

-0.59

 Promise

-0.57

rog

-0.56

ocide

-0.56

[]

-0.56

POSITIVE LOGITS

 subscrib

0.84

 subscribed

0.76

 minded

0.75

 yourself

0.75

 reading

0.74

hin

0.73

 bothered

0.73

 paying

0.71

epad

0.71

 inclined

0.70

Activations Density 0.144%

phrases indicating recommendations such as "have seen it yet," "haven't heard of," or "highly recommended."

No Comments

No Known Activations