INDEX

Explanations

words associated with news coverage and reporting

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs49152-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

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

xus

-0.77

EStream

-0.72

lain

-0.67

 ceilings

-0.65

 barriers

-0.65

 sidx

-0.65

Neh

-0.63

 aids

-0.62

 imped

-0.62

 boycot

-0.61

POSITIVE LOGITS

haus

0.82

angled

0.79

rounder

0.76

ork

0.75

arten

0.72

asia

0.70

round

0.69

ark

0.69

ername

0.68

aper

0.68

Activations Density 0.015%

words associated with news coverage and reporting

No Comments

No Known Activations