INDEX

Explanations

names or terms related to news events or legal/criminal issues

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

ï¸ı

-0.76

MRI

-0.65

raising

-0.62

 sidx

-0.60

IDER

-0.60

 Ferry

-0.60

irez

-0.60

gerald

-0.59

hillary

-0.59

-0.58

POSITIVE LOGITS

zsche

0.78

ensical

0.77

ukong

0.69

ilus

0.67

theless

0.67

vana

0.61

onsense

0.59

ional

0.59

 Slip

0.58

atural

0.58

Activations Density 1.663%

names or terms related to news events or legal/criminal issues

No Comments

No Known Activations