INDEX

Explanations

terms related to death or intense 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_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

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

MN

-0.79

umar

-0.79

Cola

-0.78

CN

-0.76

Americ

-0.74

ECA

-0.72

EEK

-0.68

Avg

-0.68

Atl

-0.68

soType

-0.67

POSITIVE LOGITS

blow

1.01

 toll

0.99

locked

0.93

stroke

0.92

guard

0.92

match

0.91

bed

0.91

adder

0.90

trap

0.89

 penalty

0.88

Activations Density 0.049%

terms related to death or intense negative consequences

No Comments

No Known Activations