INDEX

Explanations

names, specifically those starting with "Dan."

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

 overs

-0.61

 Trinidad

-0.57

 instrument

-0.57

ĵĺ

-0.55

ECB

-0.55

 footing

-0.54

rs

-0.54

 disson

-0.54

 violet

-0.53

ãģį

-0.53

POSITIVE LOGITS

riot

0.98

nery

0.97

ulas

0.97

roman

0.95

riots

0.92

afort

0.92

esian

0.91

zl

0.91

atro

0.90

roach

0.90

Activations Density 0.814%

names, specifically those starting with "Dan."

No Comments

No Known Activations