INDEX

Explanations

mentions of specific dates or time frames

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

attends to the word "between" from paired nouns.

oai_attention-head · gemini-1.5-flash Triggered by @johnny

attends to tokens indicating relationships or interactions from tokens related to conflict or comparison.

oai_attention-head · gpt-4o-mini Triggered by @johnny

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-att-kk

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L0_Hcat_z_lr1.20e-03_l11.80e+00_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.0.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

0:0.02

1:0.21

2:0.02

3:0.20

4:0.04

5:0.13

6:0.02

7:0.05

8:0.03

9:0.10

10:0.06

11:0.06

Negative Logits

IFIED

-3.00

ek

-3.00

ifer

-2.84

Mich

-2.84

ATH

-2.80

PD

-2.78

VI

-2.75

OGR

-2.72

MD

-2.65

Cost

-2.64

POSITIVE LOGITS

��

4.33

etheless

4.21

lihood

3.72

terness

3.67

imity

3.51

terday

3.37

 suspic

2.96

PsyNetMessage

2.90

htaking

2.89

ertodd

2.83

Activations Density 0.013%

mentions of specific dates or time frames

attends to the word "between" from paired nouns.

attends to tokens indicating relationships or interactions from tokens related to conflict or comparison.

No Comments

No Known Activations