INDEX

Explanations

verbs related to causing or forcing actions, often with negative consequences

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_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

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

yd

-0.70

chool

-0.64

 Birthday

-0.62

ta

-0.59

 submitted

-0.59

 Basics

-0.59

at

-0.58

 styled

-0.58

bid

-0.58

 supplied

-0.57

POSITIVE LOGITS

 havoc

1.05

 morale

0.89

 untold

0.88

 confusion

0.84

us

0.80

 resentment

0.79

 headaches

0.76

 viewers

0.75

 friction

0.74

 irre

0.74

Activations Density 5.308%

verbs related to causing or forcing actions, often with negative consequences

No Comments

No Known Activations