INDEX

Explanations

proper nouns related to locations or names, specifically focusing on "Wood" as seen in multiple activations

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_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

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

ional

-0.83

 disenfranch

-0.74

uates

-0.73

 aver

-0.72

uated

-0.70

 Uriel

-0.67

 delinqu

-0.66

 disguise

-0.66

BILITY

-0.66

uate

-0.65

POSITIVE LOGITS

ruff

1.38

stock

1.15

bridge

1.12

lands

1.12

working

1.12

lake

1.11

cock

1.10

leigh

1.08

cutting

1.07

worker

1.05

Activations Density 0.015%

proper nouns related to locations or names, specifically focusing on "Wood" as seen in multiple activations

No Comments

No Known Activations