INDEX

Explanations

time-related durations, specifically in unit of minutes and hours

oai_token-act-pair · gpt-3.5-turbo

references to durations of time, particularly those expressed in minutes or hours

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

ename

-0.71

ENC

-0.70

govtrack

-0.68

 Reborn

-0.67

 taxed

-0.66

hess

-0.65

wolves

-0.65

ettings

-0.65

ashtra

-0.64

Protect

-0.63

POSITIVE LOGITS

 intervals

0.76

 increments

0.74

istance

0.72

iversary

0.70

strument

0.69

 affair

0.68

 halftime

0.68

 version

0.66

legged

0.64

 acre

0.64

Activations Density 0.035%

time-related durations, specifically in unit of minutes and hours

references to durations of time, particularly those expressed in minutes or hours

No Comments

No Known Activations