INDEX

Explanations

phrases where something is mentioned or referred to, providing context or additional information

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

istries

-0.85

abases

-0.76

earable

-0.74

rontal

-0.73

pan

-0.71

robe

-0.70

ococ

-0.69

ogram

-0.68

emate

-0.68

ograms

-0.68

POSITIVE LOGITS

 earlier

0.89

 above

0.83

 wont

0.78

 previously

0.78

 commenters

0.77

 yesterday

0.76

 elsewhere

0.76

 aforementioned

0.75

 evidenced

0.74

 afore

0.74

Activations Density 0.996%

phrases where something is mentioned or referred to, providing context or additional information

No Comments

No Known Activations