INDEX

Explanations

news-related information, publication names, and editorial content

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

ruary

-0.72

 waivers

-0.72

 opponents

-0.71

ierrez

-0.66

 affili

-0.66

TN

-0.65

itionally

-0.65

 mitigation

-0.62

 condemnation

-0.62

 partnerships

-0.60

POSITIVE LOGITS

 Awakens

0.94

 Trilogy

0.87

 Below

0.86

 Avenger

0.83

 Collection

0.82

 Shed

0.81

 Bride

0.81

 Brigade

0.81

 Incident

0.81

Fix

0.81

Activations Density 0.979%

news-related information, publication names, and editorial content

No Comments

No Known Activations