INDEX

Explanations

This attention head attends to punctuation marks (particularly colons) from tokens that follow the colon.

oai_attention-head · claude-3-7-sonnet-20250219 Triggered by @trainjumperx

references to competitive gaming events and tournaments

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-att-kk

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L3_Hcat_z_lr1.20e-03_l19.00e-01_ds24576_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v9.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

float32

Hook Name

blocks.3.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.03

1:0.06

2:0.35

3:0.07

4:0.03

5:0.07

6:0.04

7:0.05

8:0.12

9:0.04

10:0.05

11:0.03

Negative Logits

 destro

-1.42

 contrace

-1.39

 subsequ

-1.26

 outwe

-1.24

egal

-1.20

 looph

-1.20

 suscept

-1.16

uyomi

-1.14

obe

-1.11

 centrif

-1.08

POSITIVE LOGITS

ACTED

1.59

asma

1.40

Unknown

1.33

Interstitial

1.31

FK

1.25

FU

1.25

unknown

1.25

 Conflict

1.17

ixel

1.15

lement

1.13

Activations Density 0.252%

This attention head attends to punctuation marks (particularly colons) from tokens that follow the colon.

references to competitive gaming events and tournaments

No Comments

No Known Activations