INDEX

Explanations

punctuation marks and certain character strings, indicating a focus on sentence endings and structuring in the text

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L9_Hcat_z_lr1.20e-03_l11.20e+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.9.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Head Attr Weights

0:0.05

1:0.12

2:0.06

3:0.04

4:0.06

5:0.05

6:0.08

7:0.05

8:0.05

9:0.27

10:0.03

11:0.10

Negative Logits

 Schne

-3.69

 Ellison

-3.62

****

-3.57

 Cheney

-3.56

Dex

-3.31

 Labyrinth

-3.19

rax

-3.19

 Dexter

-3.16

lock

-3.13

-3.08

POSITIVE LOGITS

 targ

2.97

 reciprocal

2.93

 elig

2.92

 resultant

2.91

 tomat

2.87

 reciproc

2.84

 foliage

2.81

 lett

2.75

bot

2.66

 resulting

2.66

Activations Density 0.002%

punctuation marks and certain character strings, indicating a focus on sentence endings and structuring in the text

No Known Activations