INDEX

Explanations

legalese and news-related phrases containing specific dates and times

oai_token-act-pair · gpt-3.5-turbo

references to specific times and events in a narrative context

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

Years

-0.72

 disadvantages

-0.70

 nurture

-0.69

 nowadays

-0.69

 dependence

-0.64

models

-0.64

 someday

-0.63

 traits

-0.63

 reliance

-0.62

 mould

-0.62

POSITIVE LOGITS

 livestream

0.97

 noon

0.92

 adjourn

0.90

 countdown

0.88

 woke

0.87

noon

0.87

morning

0.87

 midnight

0.87

 MSNBC

0.87

 afternoon

0.86

Activations Density 1.135%

legalese and news-related phrases containing specific dates and times

references to specific times and events in a narrative context

No Comments

No Known Activations