正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

currently somewhere between self improvement and self destruction...

𝒵𝒾𝓀✯

11,556 subscribers

18,345 次观看 • 11 天前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

Self improvement guide 🏆 ⚡

Self improvement guide 🏆 ⚡

Fitnesstime

68,683 次观看 • 7 个月前

Guys I think self improvement has activated

Guys I think self improvement has activated

King II: Giant Swing Boogaloo

448,355 次观看 • 1 年前

Bro’s on Self Destruction Mode😭💀

Bro’s on Self Destruction Mode😭💀

rareindianclips

64,101 次观看 • 4 个月前

There is no relationship between self-esteem and excellence.

There is no relationship between self-esteem and excellence.

Shane Parrish

20,308 次观看 • 1 个月前

Allora 🤝 Talus We’re collaborating with Allora to build self-improving AI smart agents. Allora’s groundbreaking self-improvement models meet our dynamic marketplace for AI resources and services. Read more:

Allora 🤝 Talus We’re collaborating with Allora to build self-improving AI smart agents. Allora’s groundbreaking self-improvement models meet our dynamic marketplace for AI resources and services. Read more:

Talus 🐸

52,840 次观看 • 2 年前

What if you kept asking an LLM to "make it better"? In some recent work at FAIR, we investigate how we can efficiently use RL to fine-tune LLMs to iteratively self-improve on their previous solutions at inference-time. Training for iterated self-improvement can be costly. The naive approach to training for K self-improvement steps leads to K times the number of rollout steps per episode. We introduce Exploratory Iteration (ExIt), an RL-based automatic curriculum method that bootstraps diverse training distributions of self-improvement tasks by upcycling the LLM's own responses at previous turns as the starting points for both self-improvement and *self-divergence.* In order to decide what task to train on next, the curriculum prioritizes sampling of partial turn histories that led to higher return variance in its GRPO group (a learnability score that comes for free). This automatic curriculum over the bootstrapped task space teaches the model how to perform iterated self-improvement while only ever training the model on single-step self-improvement tasks. We look at ExIt's impact in both single-turn (contest math problems) and multi-turn (BFCLv3 multi-turn tasks), as well as MLE-bench, where the LLM is run in a search scaffold to produce solutions to real Kaggle competitions. Across these eval settings, we find ExIt produces models with greater capacity for inference-time self-improvement compared to GRPO. Notably, ExIt models can self-improve on test tasks for many more steps than the typical solution depth encountered during training, including a 22% improvement in MLE-bench performance compared to GRPO.

What if you kept asking an LLM to "make it better"? In some recent work at FAIR, we investigate how we can efficiently use RL to fine-tune LLMs to iteratively self-improve on their previous solutions at inference-time. Training for iterated self-improvement can be costly. The naive approach to training for K self-improvement steps leads to K times the number of rollout steps per episode. We introduce Exploratory Iteration (ExIt), an RL-based automatic curriculum method that bootstraps diverse training distributions of self-improvement tasks by upcycling the LLM's own responses at previous turns as the starting points for both self-improvement and self-divergence. In order to decide what task to train on next, the curriculum prioritizes sampling of partial turn histories that led to higher return variance in its GRPO group (a learnability score that comes for free). This automatic curriculum over the bootstrapped task space teaches the model how to perform iterated self-improvement while only ever training the model on single-step self-improvement tasks. We look at ExIt's impact in both single-turn (contest math problems) and multi-turn (BFCLv3 multi-turn tasks), as well as MLE-bench, where the LLM is run in a search scaffold to produce solutions to real Kaggle competitions. Across these eval settings, we find ExIt produces models with greater capacity for inference-time self-improvement compared to GRPO. Notably, ExIt models can self-improve on test tasks for many more steps than the typical solution depth encountered during training, including a 22% improvement in MLE-bench performance compared to GRPO.

Minqi Jiang

41,154 次观看 • 1 年前

Sensitive content

"I advocate for the destruction and for the annihilation of the traditional family." Brand ambassador for Doritos and self-described pedophiIe Samantha Hudson

End Wokeness

2,251,864 次观看 • 2 年前

Between a self employed lady and a married man who refused to pay for her service.

Between a self employed lady and a married man who refused to pay for her service.

YabaLeftOnline

371,193 次观看 • 1 年前

Your whole wallet should not depend on twelve words hidden somewhere at home. Bron removes the seed phrase and gives you recoverable self-custody.

Your whole wallet should not depend on twelve words hidden somewhere at home. Bron removes the seed phrase and gives you recoverable self-custody.

Bron Wallet

5,046,860 次观看 • 4 个月前

HAEGEUM is a fully Korean rap song — self-produced, self-written, and self-composed — and it deserves to hit 1B before any wannabe rapper!

HAEGEUM is a fully Korean rap song — self-produced, self-written, and self-composed — and it deserves to hit 1B before any wannabe rapper!

Edi is seeing bts

11,306 次观看 • 8 个月前

how it feels to be self-aware and still repeat self-destructive patterns

how it feels to be self-aware and still repeat self-destructive patterns

miilatoooo

446,843 次观看 • 5 个月前

How it feels to be self aware and still repeat self destructive patterns

How it feels to be self aware and still repeat self destructive patterns

le.hl

205,550 次观看 • 7 个月前

imagine getting into a self driving car drunk and not using self driving

imagine getting into a self driving car drunk and not using self driving

ΛI DRIVR

566,151 次观看 • 6 个月前

Obvioisly the human race is in love with self-destruction, we are only satisfying a consumer need - Oderus Urungus

Obvioisly the human race is in love with self-destruction, we are only satisfying a consumer need - Oderus Urungus

Ash Stabbath

14,948 次观看 • 2 个月前

Army Chief Asim Munir trying to attract foreign direct investment by tear gassing Islamabad ? Great strategy for self destruction.

Army Chief Asim Munir trying to attract foreign direct investment by tear gassing Islamabad ? Great strategy for self destruction.

Salman Ahmad

36,382 次观看 • 2 年前

The building blocks of becoming your best self is found in your day to day actions. If you can look from within and be honest with yourself..you will figure out exactly what needs to be done. Self improvement is the most underrated action you can take to get ahead.

The building blocks of becoming your best self is found in your day to day actions. If you can look from within and be honest with yourself..you will figure out exactly what needs to be done. Self improvement is the most underrated action you can take to get ahead.

Miranda Cohen

58,141 次观看 • 3 年前

Somewhere between prince and angel

Somewhere between prince and angel

Popi ⛓️🌸🇦🇷SEEING STRAY KIDS 14/09🇦🇷

68,977 次观看 • 1 年前