INDEX

Explanations

names and titles of people and organizations

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

FUL

-0.72

PsyNetMessage

-0.67

ãĤ·ãĥ£

-0.60

 SERV

-0.60

 Authorization

-0.60

 Jefferson

-0.58

theless

-0.57

 Wonderland

-0.56

anchester

-0.54

inea

-0.54

POSITIVE LOGITS

rir

0.72

antage

0.70

uv

0.66

ulhu

0.64

ynski

0.63

alion

0.61

yip

0.61

hair

0.60

 supra

0.60

nels

0.60

Activations Density 0.247%

names and titles of people and organizations

No Comments

No Known Activations