INDEX

Explanations

phrases related to investigations, probes, oversight, and reviews

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

emonic

-0.70

licks

-0.68

birth

-0.67

aez

-0.66

ãĤª

-0.65

cair

-0.63

 goodbye

-0.62

yip

-0.59

own

-0.58

 maiden

-0.58

POSITIVE LOGITS

 probing

1.10

 investigation

0.99

 probe

0.98

 investigating

0.97

 Investigation

0.97

 Probe

0.96

 investigations

0.94

 inquiry

0.93

 investigates

0.92

 subpoen

0.91

Activations Density 0.674%

phrases related to investigations, probes, oversight, and reviews

No Comments

No Known Activations