INDEX

Explanations

legal terms and crime-related words

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

bler

-0.71

natureconservancy

-0.65

ered

-0.64

Fra

-0.60

 Immunity

-0.58

FACE

-0.57

Wr

-0.57

fighting

-0.56

Rose

-0.56

âĸ¬

-0.56

POSITIVE LOGITS

atively

0.89

hiba

0.89

atell

0.87

imus

0.85

ometric

0.84

ikawa

0.84

oshenko

0.84

orrow

0.83

ocom

0.82

atis

0.82

Activations Density 2.513%

legal terms and crime-related words

No Comments

No Known Activations