Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
文献紹介: Attention is not Explanation
Search
Yumeto Inaoka
March 19, 2019
Research
0
490
文献紹介: Attention is not Explanation
Yumeto Inaoka
March 19, 2019
Tweet
Share
More Decks by Yumeto Inaoka
See All by Yumeto Inaoka
文献紹介: Quantity doesn’t buy quality syntax with neural language models
yumeto
1
190
文献紹介: Open Domain Web Keyphrase Extraction Beyond Language Modeling
yumeto
0
240
文献紹介: Self-Supervised_Neural_Machine_Translation
yumeto
0
160
文献紹介: Comparing and Developing Tools to Measure the Readability of Domain-Specific Texts
yumeto
0
170
文献紹介: PAWS: Paraphrase Adversaries from Word Scrambling
yumeto
0
160
文献紹介: Beyond BLEU: Training Neural Machine Translation with Semantic Similarity
yumeto
0
280
文献紹介: EditNTS: An Neural Programmer-Interpreter Model for Sentence Simplification through Explicit Editing
yumeto
0
340
文献紹介: Decomposable Neural Paraphrase Generation
yumeto
0
230
文献紹介: Analyzing the Limitations of Cross-lingual Word Embedding Mappings
yumeto
0
230
Other Decks in Research
See All in Research
[輪講] SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features
nk35jk
2
950
20250502_ABEJA_論文読み会_スライド
flatton
0
200
EOGS: Gaussian Splatting for Efficient Satellite Image Photogrammetry
satai
4
490
SNLP2025:Can Language Models Reason about Individualistic Human Values and Preferences?
yukizenimoto
0
110
[CV勉強会@関東 CVPR2025] VLM自動運転model S4-Driver
shinkyoto
2
480
まずはここから:Overleaf共同執筆・CopilotでAIコーディング入門・Codespacesで独立環境
matsui_528
2
460
【輪講資料】Moshi: a speech-text foundation model for real-time dialogue
hpprc
3
640
不確実性下における目的と手段の統合的探索に向けた連続腕バンディットの応用 / iot70_gp_rff_mab
monochromegane
2
150
Galileo: Learning Global & Local Features of Many Remote Sensing Modalities
satai
3
220
Creation and environmental applications of 15-year daily inundation and vegetation maps for Siberia by integrating satellite and meteorological datasets
satai
3
260
AlphaEarth Foundations: An embedding field model for accurate and efficient global mapping from sparse label data
satai
1
180
Google Agent Development Kit (ADK) 入門 🚀
mickey_kubo
2
1.7k
Featured
See All Featured
BBQ
matthewcrist
89
9.8k
Principles of Awesome APIs and How to Build Them.
keavy
126
17k
Refactoring Trust on Your Teams (GOTO; Chicago 2020)
rmw
34
3.1k
Building Adaptive Systems
keathley
43
2.7k
I Don’t Have Time: Getting Over the Fear to Launch Your Podcast
jcasabona
33
2.4k
Producing Creativity
orderedlist
PRO
347
40k
YesSQL, Process and Tooling at Scale
rocio
173
14k
Navigating Team Friction
lara
189
15k
Measuring & Analyzing Core Web Vitals
bluesmoon
9
580
What’s in a name? Adding method to the madness
productmarketing
PRO
23
3.6k
Reflections from 52 weeks, 52 projects
jeffersonlam
352
21k
Documentation Writing (for coders)
carmenintech
74
5k
Transcript
Attention is not Explanation 文献紹介 2019/03/19 長岡技術科学大学 自然言語処理研究室 稲岡 夢人
Literature 2 Title Attention is not Explanation Author Sarthak Jain,
Byron C. Wallace Conference NAACL-HLT 2019 Paper https://arxiv.org/abs/1902.10186
Abstract Attentionは入力単位で重みを示す その分布が重要性を示すものとして扱われることがある → 重みと出力の間の関係は明らかにされていない 標準的なAttentionは意味のある説明を提供しておらず、 そのように扱われるべきでないことを示す
3
調査方法 1. 重みが重要度と相関しているか 2. 事実と反する重み設定が予測を変化させるか 4
Tasks Binary Text Classification Question Answering (QA)
Natural Language Inference (NLI) 5
Datasets 6
Results 7
Definitions 出力結果の比較に使用する距離 Attentionの比較に使用する距離 8
Results Attentionの変化を大きくしても結果が変化しない → 出力への影響が小さい 9
Results DiabetesはPositiveのクラスにおいては影響が大きい → 高精度で糖尿病を示すトークンが存在するため 10
Adversarial Attention 出力を大きく変化させるようにAttentionを変化させる Attentionが少し変化しただけで出力が大きく変化するか ← Attentionの挙動を確認 12
Results 少しのAttentionの変化で出力が大きく変化している 13
Conclusions 重要度とAttentionの重みは相関が弱い 事実に反する重みは必ずしも出力を変化させない Seq2seqについては今後の課題とする 14