INDEX

Explanations

phrases related to safety or protection

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

quart

-0.81

lished

-0.76

RESULTS

-0.76

licks

-0.75

avez

-0.74

onde

-0.73

aminer

-0.73

ruary

-0.72

zig

-0.71

gres

-0.71

POSITIVE LOGITS

 shield

0.89

 Protective

0.85

 custody

0.82

 protective

0.82

 enclosure

0.79

 apparatus

0.78

 casing

0.78

 encl

0.77

 coat

0.76

 duty

0.76

Activations Density 0.009%

phrases related to safety or protection

No Comments

No Known Activations