INDEX

Explanations

names or titles preceded by "Mr."

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

 appre

-0.85

 actionGroup

-0.79

 destro

-0.72

 wheelchair

-0.72

 tremend

-0.70

rawdownloadcloneembedreportprint

-0.67

phia

-0.66

pmwiki

-0.64

orph

-0.64

 euth

-0.63

POSITIVE LOGITS

 Hyde

0.87

 Claus

0.83

 Freeze

0.80

 Universe

0.79

 Obama

0.79

 Hussein

0.79

 Farage

0.77

Ack

0.76

 Justice

0.75

 Putin

0.74

Activations Density 0.248%

names or titles preceded by "Mr."

No Comments

No Known Activations