INDEX

Explanations

terms associated with serious harm or injury

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs49152-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

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

bye

-0.73

LCS

-0.61

Eag

-0.60

 Tanz

-0.58

 foam

-0.58

 Principal

-0.58

 Finn

-0.56

 blindly

-0.55

 Noise

-0.55

 electronically

-0.55

POSITIVE LOGITS

ous

2.73

ously

2.47

OUS

1.57

osity

1.38

iously

1.34

istically

1.30

istic

1.28

ized

1.22

ious

1.22

izing

1.22

Activations Density 0.018%

terms associated with serious harm or injury

No Comments

No Known Activations