INDEX

Explanations

references to geographical locations, specifically related to North Korea and North Carolina, as well as legal and governmental terms

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_fs1536-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_1536

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

1,536

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

 "$:/

-0.82

TED

-0.76

NRS

-0.70

ATURE

-0.69

-0.67

TABLE

-0.66

wcsstore

-0.66

inventoryQuantity

-0.65

ILA

-0.65

ername

-0.64

POSITIVE LOGITS

ward

1.22

ampton

1.20

western

1.20

 Carolina

1.19

side

1.14

 Korea

1.09

bound

1.07

shore

1.06

 Dakota

1.05

 Pole

1.04

Activations Density 0.969%

references to geographical locations, specifically related to North Korea and North Carolina, as well as legal and governmental terms

No Comments

No Known Activations