Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Sign up for free
Menu
Search
Features
All features
Private URLs
Password Protection
Custom URLS
Scheduled publishing
Remove Branding
Restrict embedding
Deck Collections
Notes
Features
All features
Private URLs
Password Protection
Custom URLS
Scheduled publishing
Remove Branding
Restrict embedding
Deck Collections
Notes
Explore
Featured decks
Featured speakers
Programming
Technology
Storyboards
Explore
Featured decks
Featured speakers
Programming
Technology
Storyboards
Pricing
Search
Sign in
Sign up for free
CoCon
Search
Zhang Yixiao
December 16, 2020
Science
430
0
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
CoCon
Zhang Yixiao
December 16, 2020
More Decks by Zhang Yixiao
See All by Zhang Yixiao
vq-cpc
ldzhangyx
0
420
MixPoet
ldzhangyx
4
480
diora
ldzhangyx
0
330
drummernet
ldzhangyx
0
290
ON-LSTM
ldzhangyx
0
260
Other Decks in Science
See All in Science
Anomalies de terrain remontées à l'ONAB
institutdelelevage
PRO
0
300
Leitner Inauguration Lecture Chalmers University of Technology
xleitix
0
390
生成AI時代のインタラクションサイエンス(招待講演)/interaction-science
okana2ki
0
110
Massey Ratings for Match Outcome Prediction in Table Tennis: Evidence of Greater Stability than the ITTF World Ranking
konakalab
0
180
ssmonline #51 ヤマサキ春のサメ祭り 2026 / ssmjp Yamasaki Spring JAWS Festival 2026
naospon
1
170
機械学習 - pandas入門
trycycle
PRO
0
780
Understanding CVP Waveforms: Interpretation and Clinical Implications in Anesthesiology
taka88
0
980
Build your own LLM, Live, with MicroGPT
ianozsvald
0
160
コーヒー豆様核 (Coffee-bean nuclei) における形態学的サブタイピングと精選・焙煎特性の同定
jagupath
PRO
0
220
医療 LLM ベンチマークの現在地:多面的評価 と日本ローカライズ
analokmaus
1
760
JSAI2026企画セッションKS-14 インタビュー集『⼈⼯知能と哲学と四つの問い』が提起する⼈⼯知能のこれからの課題 趣旨説明 / JSAI2026 Special Session: A Collection of Interviews, “Artificial Intelligence, Philosophy, and Four Questions”
ykiyota
0
470
Cross-Media Technologies, Information Science and Human-Information Interaction
signer
PRO
3
33k
Featured
See All Featured
Mozcon NYC 2025: Stop Losing SEO Traffic
samtorres
1
580
Information Architects: The Missing Link in Design Systems
soysaucechin
1
1.2k
The Director’s Chair: Orchestrating AI for Truly Effective Learning
tmiket
1
330
The AI Revolution Will Not Be Monopolized: How open-source beats economies of scale, even for LLMs
inesmontani
PRO
3
3.8k
How to build a perfect <img>
jonoalderson
1
6.1k
Reality Check: Gamification 10 Years Later
codingconduct
0
2.3k
Intergalactic Javascript Robots from Outer Space
tanoku
273
28k
The Art of Delivering Value - GDevCon NA Keynote
reverentgeek
16
2.2k
Responsive Adventures: Dirty Tricks From The Dark Corners of Front-End
smashingmag
254
22k
Why Mistakes Are the Best Teachers: Turning Failure into a Pathway for Growth
auna
0
330
The browser strikes back
jonoalderson
0
1.8k
We Analyzed 250 Million AI Search Results: Here's What I Found
joshbly
2
2.1k
Transcript
CoCon: A Self-Supervised Approach for Controlled Text Generation Presenter: Yixiao
Zhang
TL;DR • 任务:用文本指导文本,进行可控生成 • 亮点: • 让文本成为控制变量:更加灵活 • 漂亮的损失函数 •
模型是自监督训练的 • 结果表明显著加强了语言模型的可控性
Introduction • 基于Transformer的预训练LM成为了新的浪潮,但是从头训练LM 的成本巨大 • 问题:不改变预训练LM的情况下,LM如何进行控制? • 解决办法1:PPLM,通过属性模型控制生成文本 • 缺点:不够精细,可能造成巨大差异
Sumanth Dathathri, Andrea Madotto, Janice Lan, Jane Hung, Eric Frank, Piero Molino, Jason Yosinski, and Rosanne Liu. Plug and play language models: a simple approach to controlled text generation. arXiv preprint arXiv:1912.02164, 2019.
Related Work • 生成包含所需属性的文本 • 早期工作 • 条件生成模型,可通过RL或GAN训练 • 缺陷:对预定属性的要求限制了生成文本的可能类型
• CTRL • 使用control code(预置的metadata)生成文本 • 缺陷:control code也是预先设定的 • PPLM(最相似) • 在LM上插拔一个模块,不重新训练实现生成 • 区别: • 本文旨在在更局部的内容上控制 • CoCon自监督学习,免去了标签数据
Related Work • 文本风格迁移 • 少数研究采用AE以分离表示 • 另外一些模型能识别attribute markers •
一些特定风格相关的n-grams • 通过替换的方式编辑文本风格
CoCon • 模型目标: • 给定引导文本1:−1 和控制文本,模型 生成: • 过程: •
分别编码c和x • 自注意力交互,得 到新的特征 • 进行下一个词预测
CoCon • CoCon是一个单层 Transformer Block • 首先得到x和c的QKV • 将KV拼起来过self-attn
Loss Function • 自重构损失(Self Reconstruction Loss) • 令c = ,使得模型能够学习结合控制文本的内容
• 无文本损失(Null Content Loss) • 令 = ∅,使得模型退化成LM,以生成流畅的文本
Loss Function • 循环重构损失(Cycle Reconstruction Loss) • 在inference中,生成文本不太可能与引导文本共存 • 给定两个不同的文本(,
′)
Loss Function • 对抗损失(Adversarial Loss) • 总优化目标
实验 • 文本引导的文本生成评估指标有BLEU、NIST、METEOR、PPL和 Dist-1/2/3
实验 • 文本引导的文本生成评估指标有BLEU、NIST、METEOR、PPL和 Dist-1/2/3
例子
多个控制