INDEX

Explanations

politically related terms and news

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_fs12288-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_12288

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

12,288

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

efer

-0.76

cium

-0.76

ulative

-0.76

iphate

-0.75

ses

-0.72

sheet

-0.72

iak

-0.70

ombat

-0.69

ocaust

-0.69

verts

-0.69

POSITIVE LOGITS

 unsurprisingly

1.16

 someday

0.84

 unsur

0.84

 sensing

0.83

haps

0.81

 surprisingly

0.77

 exacerbated

0.76

 embold

0.74

 opio

0.73

 indicative

0.73

Activations Density 0.084%

politically related terms and news

No Comments

No Known Activations