INDEX

Explanations

information related to news articles or journalistic reports

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L2_Hcat_z_lr1.20e-03_l11.00e+00_ds24576_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v4.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

float32

Hook Name

blocks.2.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Head Attr Weights

0:0.04

1:0.03

2:0.16

3:0.08

4:0.10

5:0.09

6:0.09

7:0.02

8:0.15

9:0.09

10:0.06

11:0.03

Negative Logits

pora

-1.17

spir

-1.14

afety

-1.14

ILA

-1.12

birds

-1.08

uana

-1.05

communications

-1.03

 mosqu

-1.02

 FANTASY

-1.02

liest

-1.01

POSITIVE LOGITS

 guiActiveUnfocused

1.22

 targ

1.18

daq

1.18

legram

1.12

cue

1.10

Cy

1.07

verett

1.07

 Cosponsors

1.07

oqu

1.04

 sidx

1.03

Activations Density 0.002%

information related to news articles or journalistic reports

No Known Activations