INDEX

Explanations

proper names or terms prefixed with "El"

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

hire

-0.65

olicy

-0.63

 raft

-0.63

 charge

-0.61

 Eliot

-0.60

henko

-0.60

nant

-0.59

LESS

-0.59

ratulations

-0.59

ELS

-0.58

POSITIVE LOGITS

 Paso

1.19

abor

1.14

igible

1.08

 Salvador

1.01

usive

1.00

iza

0.99

 NiÃ±o

0.99

azer

0.96

apsed

0.93

isa

0.90

Activations Density 0.017%

proper names or terms prefixed with "El"

No Comments

No Known Activations