INDEX

Explanations

terms related to reports or analyses of diverse subjects, potentially aiming at identification or assessment

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

marine

-0.69

 âĢ¦"

-0.61

itiz

-0.59

 ..."

-0.58

igers

-0.58

 situ

-0.54

 doorstep

-0.54

Eat

-0.54

 farm

-0.53

 either

-0.53

POSITIVE LOGITS

entimes

0.79

ward

0.77

 hindsight

0.75

bestos

0.75

math

0.75

itialized

0.75

cknowled

0.74

 contrast

0.72

 meantime

0.72

nce

0.71

Activations Density 3.392%

terms related to reports or analyses of diverse subjects, potentially aiming at identification or assessment

No Comments

No Known Activations