INDEX

Explanations

highly negative or controversial statements/actions in text related to schools, race, international conflicts, and political figures

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

RAFT

-0.78

rex

-0.67

romy

-0.63

rn

-0.63

midt

-0.61

staking

-0.61

iership

-0.61

romeda

-0.61

(){

-0.61

ean

-0.60

POSITIVE LOGITS

xious

1.31

 longer

1.10

 doubt

1.07

 matter

1.01

except

1.00

 shortage

0.94

ct

0.94

 indication

0.89

obs

0.87

 discern

0.86

Activations Density 0.831%

highly negative or controversial statements/actions in text related to schools, race, international conflicts, and political figures

No Comments

No Known Activations