INDEX

Explanations

financial terms, particularly related to news and reports from specific sources like "WS" and "J"

oai_token-act-pair · gpt-3.5-turbo

references to a specific news outlet or website

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.5.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

ments

-0.85

staking

-0.83

aneous

-0.82

iasis

-0.75

 Kurd

-0.71

rical

-0.71

aneously

-0.70

inant

-0.69

iating

-0.69

iations

-0.68

POSITIVE LOGITS

ullivan

1.02

nesday

0.92

alez

0.91

kins

0.90

nyder

0.89

ocket

0.85

erver

0.82

atcher

0.81

addle

0.81

WS

0.80

Activations Density 0.015%

financial terms, particularly related to news and reports from specific sources like "WS" and "J"

references to a specific news outlet or website

No Comments

No Known Activations