INDEX

Explanations

words related to police, government, corruption, and investigations

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

cients

-0.70

soever

-0.67

issors

-0.65

worker

-0.61

sson

-0.60

Gum

-0.57

Dro

-0.57

ktop

-0.56

 Kenobi

-0.56

 Unlimited

-0.56

POSITIVE LOGITS

 intensity

0.97

intensity

0.90

priority

0.90

 ante

0.88

 levels

0.87

profile

0.85

 sophistication

0.83

Grade

0.82

 stakes

0.81

ensity

0.80

Activations Density 1.087%

words related to police, government, corruption, and investigations

No Comments

No Known Activations