INDEX

Explanations

terms related to controversial or sensitive legal and ethical issues

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

ICES

-0.75

 embassies

-0.72

 avenues

-0.71

doms

-0.70

 lobb

-0.70

 workshops

-0.69

nces

-0.69

 timelines

-0.69

places

-0.68

 testimonies

-0.67

POSITIVE LOGITS

or

0.79

 person

0.74

 sweater

0.73

guy

0.72

assian

0.71

oscope

0.71

lander

0.70

 pillow

0.69

 whore

0.69

entary

0.69

Activations Density 0.355%

terms related to controversial or sensitive legal and ethical issues

No Comments

No Known Activations