INDEX

Explanations

keywords related to investigative analysis and cross-referencing evidence

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

ulously

-0.72

ICAN

-0.70

ILY

-0.66

FORE

-0.64

ITAL

-0.63

idity

-0.61

lain

-0.59

Vac

-0.59

ately

-0.58

ECH

-0.58

POSITIVE LOGITS

roads

1.29

bow

1.15

bows

1.06

hair

0.98

cutting

0.98

word

0.94

bones

0.93

dress

0.88

continental

0.87

wal

0.85

Activations Density 0.404%

keywords related to investigative analysis and cross-referencing evidence

No Comments

No Known Activations