Part 9 of our 12-part series covering the most recent NeurIPS Conference explores innovative approaches to combining and merging language models. This presentation covers:
Overview of evolutionary strategies and genetic algorithms for model merging
Detailed analysis of contrastive decoding techniques
Practical examples of model merging across different architectures
Deep dive into logit manipulation and probability differences
Proxy tuning methods for efficient model adaptation
Real-world applications including fact-checking and structured reasoning
This session is part of our comprehensive NeurIPS conference review series, focusing on practical implementations and real-world applications of the latest AI research.
Below is a link to the presentation deck: https://docs.google.com/presentation/...
Поделиться в:
Доступные форматы для скачивания:
Скачать видео mp4
Информация по загрузке:
Скачать аудио mp3
Похожие видео
array(10) {
[0]=>
object(stdClass)#4397 (5) {
["video_id"]=>
int(9999999)
["related_video_id"]=>
string(11) "bN_nWYc4mac"
["related_video_title"]=>
string(92) "Visual AI Breakthroughs: Next-Scale Prediction and World Models from NeurIPS (Part 10 of 12)"
["posted_time"]=>
string(25) "4 месяца назад"
["channelName"]=>
string(37) "RelationalAI - AI for Relational Data"
}
[1]=>
object(stdClass)#4370 (5) {
["video_id"]=>
int(9999999)
["related_video_id"]=>
string(11) "M7tgQ8-z26Y"
["related_video_title"]=>
string(77) "Data Quality & Training Rigor in AI: Lessons from NeurIPS (Part 11 of 12)"
["posted_time"]=>
string(25) "4 месяца назад"
["channelName"]=>
string(37) "RelationalAI - AI for Relational Data"
}
[2]=>
object(stdClass)#4395 (5) {
["video_id"]=>
int(9999999)
["related_video_id"]=>
string(11) "OPDoc4yqKgI"
["related_video_title"]=>
string(103) "LLM Scientist Toolkit: Key Insights from NeurIPS on Data Efficiency & Model Training (Part 8 of 12)"
["posted_time"]=>
string(25) "4 месяца назад"
["channelName"]=>
string(37) "RelationalAI - AI for Relational Data"
}
[3]=>
object(stdClass)#4402 (5) {
["video_id"]=>
int(9999999)
["related_video_id"]=>
string(11) "-7dDF2GDUdg"
["related_video_title"]=>
string(85) "AI Scaling Laws & Compute Tradeoffs: From Games to Language Models (Part 4 of 12)"
["posted_time"]=>
string(25) "4 месяца назад"
["channelName"]=>
string(37) "RelationalAI - AI for Relational Data"
}
[4]=>
object(stdClass)#4381 (5) {
["video_id"]=>
int(9999999)
["related_video_id"]=>
string(11) "mThiyFYEQhY"
["related_video_title"]=>
string(163) "«Будем жить!» | Хитрая передача на Первом канале о вернувшихся с СВО (English subtitles) @Max_Katz"
["posted_time"]=>
string(21) "3 часа назад"
["channelName"]=>
string(19) "Максим Кац"
}
[5]=>
object(stdClass)#4399 (5) {
["video_id"]=>
int(9999999)
["related_video_id"]=>
string(11) "iyrnPNBWIQ4"
["related_video_title"]=>
string(161) "«Жить надо сегодня». Олег Тиньков и Майкл Калви о взлете нового финтех-стартапа Plata"
["posted_time"]=>
string(21) "1 день назад"
["channelName"]=>
string(28) "Это Осетинская!"
}
[6]=>
object(stdClass)#4394 (5) {
["video_id"]=>
int(9999999)
["related_video_id"]=>
string(11) "ztrdSWvpRKM"
["related_video_title"]=>
string(59) "ПЕРВЫЙ капсульный поезд: 5000 руб.!"
["posted_time"]=>
string(21) "1 день назад"
["channelName"]=>
string(3) "808"
}
[7]=>
object(stdClass)#4404 (5) {
["video_id"]=>
int(9999999)
["related_video_id"]=>
string(11) "QFK6RG47bww"
["related_video_title"]=>
string(58) "The Factory of Ideas: Working at Bell Labs - Computerphile"
["posted_time"]=>
string(19) "9 лет назад"
["channelName"]=>
string(13) "Computerphile"
}
[8]=>
object(stdClass)#4380 (5) {
["video_id"]=>
int(9999999)
["related_video_id"]=>
string(11) "sgVosaLhPV0"
["related_video_title"]=>
string(35) "Computers Are Just Rocks Doing Math"
["posted_time"]=>
string(19) "4 дня назад"
["channelName"]=>
string(18) "The Science Asylum"
}
[9]=>
object(stdClass)#4398 (5) {
["video_id"]=>
int(9999999)
["related_video_id"]=>
string(11) "Y5vwCpMbhH4"
["related_video_title"]=>
string(83) "Data Monetization & Model Attribution in AI: Technical Deep Dive (Part 5 of 12)"
["posted_time"]=>
string(25) "4 месяца назад"
["channelName"]=>
string(37) "RelationalAI - AI for Relational Data"
}
}
