INDEX

Explanations

references to political parties and candidates

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

alam

-0.74

 Grizz

-0.73

 Kard

-0.71

 Dull

-0.70

ravings

-0.69

Hoo

-0.67

 Wheat

-0.66

Maw

-0.65

DEN

-0.65

DERR

-0.64

POSITIVE LOGITS

 affiliation

1.14

 affili

1.00

 leadership

0.93

goers

0.90

leader

0.87

Leader

0.85

 membership

0.84

 leaders

0.84

 faithful

0.83

 leader

0.82

Activations Density 0.381%

references to political parties and candidates

No Comments

No Known Activations