INDEX

Explanations

references to news reporters or media personnel

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

osate

-0.69

ez

-0.67

UTH

-0.66

ston

-0.66

inho

-0.65

tein

-0.65

UCT

-0.63

phal

-0.63

tail

-0.62

uras

-0.61

POSITIVE LOGITS

 gathered

0.90

hips

0.89

 afterward

0.88

 aboard

0.85

 reporters

0.82

 stationed

0.82

 assembled

0.79

 Thursday

0.76

hip

0.76

 afterwards

0.75

Activations Density 0.015%

references to news reporters or media personnel

No Comments

No Known Activations