INDEX

Explanations

specific names and references related to individuals, political parties or positions, and specific 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_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

»Ĵ

-0.81

å§«

-0.77

ĥ

-0.73

ĺ

-0.72

ousands

-0.70

acas

-0.69

Ħ¢

-0.69

alyses

-0.68

ĭ

-0.67

ãĥł

-0.67

POSITIVE LOGITS

 fault

0.93

 anyway

0.80

 equivalent

0.79

 rather

0.77

 minus

0.77

 versus

0.73

 consolation

0.72

 disguised

0.72

 reborn

0.71

 Incarn

0.70

Activations Density 2.716%

specific names and references related to individuals, political parties or positions, and specific events

No Comments

No Known Activations