Shunyu Yao @ShunyuYao12 Twitter profile

Last Seen Profiles

@garamluver

@official_tshaw

@DouglasBShaw

@Guepardow

@Volumpio

@Tony_Flame

@chiYuki223

@Jun1E1i

@IABEurope

@AbeAAdekunle

@mehmetipsirli

@jacques_pepin

@ChitkwesuManetu

@iamlicous

@Cheetahs_R_Life

@Mark_Wilson_

@sescenografia

@bin278032114224

@MoveDAO_

@bisstcofy

@BeachClubSolana

@Camelli34863154

@urasawa_naoki

@EmbaMexEsp

@ajitamaumaina

@TVQuase

@R7s4n_

@cvoizenet

@montserratarias

@FiercelyVirgo

@surendersihag22

@JenBott_

@usuu_to_zenbu

@IG0RSAPNAP

@AminShahab5

@JennyBotTV

Shunyu Yao

@ShunyuYao12

2 months

almost forget to post: i've joined @OpenAI ! time to convert the research vision to reality, and expect something exciting to drop out :)

83

18

1K

Shunyu Yao

@ShunyuYao12

1 year

Still use ⛓️Chain-of-Thought (CoT) for all your prompting? May be underutilizing LLM capabilities🤠 Introducing 🌲Tree-of-Thought (ToT), a framework to unleash complex & general problem solving with LLMs, through a deliberate ‘System 2’ tree search.

93

594

3K

Shunyu Yao

@ShunyuYao12

5 months

I've defended my PhD! "Language Agents: From Next-Token Prediction to Digital Automation" - Talk (WebShop, SWE-bench, ReAct, ToT, CoALA, and on the future of agents): - Thesis (covers even more):

45

64

814

Shunyu Yao

@ShunyuYao12

5 months

I will present my thesis defense tomorrow! Language Agents: From Next-Token Prediction to Digital Automation - 10am EST on Thursday, May 2 - - WebShop, ReAct, ToT, CoALA - Briefly: SWE-bench/agent - Thoughts on the future of language agents

24

54

670

Shunyu Yao

@ShunyuYao12

1 year

🧠🦾ReAct -> 🔥FireAct Most language agents prompt LMs - ReAct, AutoGPT, ToT, Generative Agents, ... - Which is expensive, slow, and non-robust😢 Most fine-tuned LMs not for agents... FireAct asks: WHY NOT? Paper, code, data, ckpts: (1/5)

6

112

483

Shunyu Yao

@ShunyuYao12

1 year

Language Agents are cool & fast-moving, but no systematic way to understand & design them.. So we use classical CogSci & AI insights to propose Cognitive Architectures for Language Agents (🐨CoALA)! w/ great @tedsumers @karthik_r_n @cocosci_lab (1/6)

10

116

464

Shunyu Yao

@ShunyuYao12

7 months

Solving >10% of our SWE-Bench () is THE most impressive result in 2024 so far, and a milestone for the research and application of AI agents. Congrats @cognition_labs !

Cognition

@cognition_labs

7 months

Today we're excited to introduce Devin, the first AI software engineer. Devin is the new state-of-the-art on the SWE-Bench coding benchmark, has successfully passed practical engineering interviews from leading AI companies, and has even completed real jobs on Upwork. Devin is

5K

11K

45K

14

37

444

Shunyu Yao

@ShunyuYao12

2 years

Large Language Models (LLM) are 🔥in 2 ways: 1.🧠Reason via internal thoughts (explain jokes, math reasoning..) 2.💪Act in external worlds (SayCan, ADEPT ACT-1, WebGPT..) But so far 🧠and💪 remain distinct methods/tasks... Why not 🧠+💪? In our new work ReAct, we show 1+1>>2!

AK

@_akhaliq

2 years

ReAct: Synergizing Reasoning and Acting in Language Models abs:

3

26

150

10

66

389

Shunyu Yao

@ShunyuYao12

1 year

The art of programming is interactive. Why should coding benchmarks be "seq2seq"? Thrilled to present 🔄InterCode, next-gen framework of coding tasks as standard RL tasks (action=code, observation=execution feedback) paper, code, data, pip: (1/7)

11

69

387

Shunyu Yao

@ShunyuYao12

1 year

Code released at , thanks for waiting! It's intentionally kept minimalistic (core ~ 100 lines), though some features (e.g. variable breadth across steps) can be easily added to improve perf & reduce cost. (1/2)

GitHub - princeton-nlp/tree-of-thought-llm: [NeurIPS 2023] Tree of Thoughts: Deliberate Problem...

[NeurIPS 2023] Tree of Thoughts: Deliberate Problem Solving with Large Language Models - princeton-nlp/tree-of-thought-llm

github.com

Shunyu Yao

@ShunyuYao12

1 year

Still use ⛓️Chain-of-Thought (CoT) for all your prompting? May be underutilizing LLM capabilities🤠 Introducing 🌲Tree-of-Thought (ToT), a framework to unleash complex & general problem solving with LLMs, through a deliberate ‘System 2’ tree search.

93

594

3K

6

73

305

Shunyu Yao

@ShunyuYao12

5 months

Coding is the frontier of AI. Excited to push the two frontiers of AI coding: 1. SWE(-bench/agent) 2. Olympiad programming (this tweet) Introduce USACO benchmark: * inference methods (RAG/reflect) help a bit: 9->20% * human feedback helps a lot: 0->86%!

8

50

303

Shunyu Yao

@ShunyuYao12

5 months

SWE-agent paper: If you care about - SWE: check detailed design & analysis - agents: check principles & insights for agent-computer interface (ACI) - HCI/CogSci: check parallels in ACI/understanding LMs! chat w/ @jyangballin @_carlosejimenez @iclr_conf !

0

29

201

Shunyu Yao

@ShunyuYao12

6 months

Extremely excited to open-source our SWE-agent that achieves SoTA on SWE-bench😃 Turns out ReAct + Agent-Computer Interface (ACI) can go a long way, very excited about the implications for SWE and beyond!

John Yang

@jyangballin

6 months

SWE-agent is our new system for autonomously solving issues in GitHub repos. It gets similar accuracy to Devin on SWE-bench, takes 93 seconds on avg + it's open source! We designed a new agent-computer interface to make it easy for GPT-4 to edit+run code

64

426

2K

5

16

166

Shunyu Yao

@ShunyuYao12

1 year

Major update: we made a pip package for ToT! pip install tree-of-thoughts-llm Learn more about how to use ToT for your use cases:

Shunyu Yao

@ShunyuYao12

1 year

Still use ⛓️Chain-of-Thought (CoT) for all your prompting? May be underutilizing LLM capabilities🤠 Introducing 🌲Tree-of-Thought (ToT), a framework to unleash complex & general problem solving with LLMs, through a deliberate ‘System 2’ tree search.

93

594

3K

3

44

168

Shunyu Yao

@ShunyuYao12

1 month

We have confirmed a list of awesome speakers and panelists for our #NeurIPS2024 workshop on System 2 Reasoning at Scale! Come on Dec 15!

3

9

136

Shunyu Yao

@ShunyuYao12

15 days

This is just the beginning of a new paradigm. So much more to come :)

OpenAI

@OpenAI

15 days

We're releasing a preview of OpenAI o1—a new series of AI models designed to spend more time thinking before they respond. These models can reason through complex tasks and solve harder problems than previous models in science, coding, and math.

985

4K

18K

4

2

122

Shunyu Yao

@ShunyuYao12

1 year

A summary thread of our recent (i.e. after GPT-3) work in language agents, in tweets👇 ( also provides a nice summary --- I might be the first researcher that includes "tweet" column for publication?🤷)

4

17

120

Shunyu Yao

@ShunyuYao12

17 days

Will talk about history and future of llm agents next monday - stay tuned!

Dawn Song

@dawnsongtweets

21 days

Large Language Model Agents is the next frontier. Really excited to announce our Berkeley course on LLM Agents, also available for anyone to join as a MOOC, starting Sep 9 (Mon) 3pm PT! 📢 Sign up & join us:

26

254

1K

1

6

115

Shunyu Yao

@ShunyuYao12

10 months

I'll give an oral talk about Tree of Thoughts @NeurIPSConf at 3:45-4pm CST on Dec 13 (4C), with the poster session right after ( #410 ). I'm also on the faculty job market this year, so DM me if you wanna chat😃 (Other posters: InterCode #522 , Reflexion #508 , 5-7pm Dec 14)

Shunyu Yao

@ShunyuYao12

1 year

Still use ⛓️Chain-of-Thought (CoT) for all your prompting? May be underutilizing LLM capabilities🤠 Introducing 🌲Tree-of-Thought (ToT), a framework to unleash complex & general problem solving with LLMs, through a deliberate ‘System 2’ tree search.

93

594

3K

2

13

100

Shunyu Yao

@ShunyuYao12

1 year

What to do if someone implemented my work (tree of thoughts) but failed to acknowledge what is the offical repo, and have more stars than offical repo, and might mislead people about the content of the work (i.e. implementation might not be paper's ideas)

Is there a way to get that other fake tree-of-thoughts repo taken down? · Issue #17 · princeton-n...

Sorry if this isn't the place to post his inquiry, but I had, at first, given up on your process while thinking this spamming troll was you guys. Obviously, it's auto-generated fake code th...

github.com

7

22

97

Shunyu Yao

@ShunyuYao12

2 months

Very excited for this convergence of interest :)

OpenAI

@OpenAI

2 months

We're releasing a new iteration of SWE-bench, in collaboration with the original authors, to more reliably evaluate AI models on their ability to solve real-world software issues.

409

522

3K

0

3

95

Shunyu Yao

@ShunyuYao12

6 months

People still surprised by such things across pairs among ReAct ToT Reflexion CoALA WebShop SWE-bench SWE-agent😂

7

3

94

Shunyu Yao

@ShunyuYao12

1 year

Write a sentence with "dog, frisbee, catch, throw" 👉Too easy for 7B LM... Will (constrained) text generation (CTG) "die out" like many other NLP tasks, in face of LLM? 👉Excited to introduce 🐕COLLIE, next-gen CTG that even challenges GPT-4! (1/n)

2

13

94

Shunyu Yao

@ShunyuYao12

1 year

Meme aside, Check SWE-bench that hits many checks for a good benchmark - hard but useful to solve, easy to evaluate - automatically constructed from real GitHub issues and pull requests - challenge super long context, retrieval, coding, etc - can easily update with new instances

carlos

@_carlosejimenez

1 year

Can LMs 🤖 replace programmers 🧑‍💻? - Not yet! Our new benchmark, SWE-bench, tests models on solving real issues from GitHub. Claude 2 & GPT-4 get <5% acc. 🔗 See our leaderboard, paper, code, data: 🧵

12

90

474

1

10

88

Shunyu Yao

@ShunyuYao12

1 year

We show huge gains on 3 new tasks GPT-4 can't solve directly or with CoT (hard to find!) due to a need for planning / searching: game of 24, creative writing, crosswords.

3

1

82

Shunyu Yao

@ShunyuYao12

1 year

Not at #ICML2023 but happy to finally release a @princeton_nlp blog post written by me and @karthik_r_n on the opportunities and risks of language agents should be a fun 10min read! it's a very new subject, so please leave any comments here👇

2

24

83

Shunyu Yao

@ShunyuYao12

7 days

Huge thanks to @dawnsongtweets @xinyun_chen_ for inviting me to such a timely, well-organized, and extremely popular class! Check out the recording for my talk on the history and overview of llm agents - hope u like it 😀

Dawn Song

@dawnsongtweets

7 days

Join us for 3rd lecture on agent frameworks in our LLM Agents MOOC, @chi_wang_ @GoogleDeepMind & @jerryjliu0 @llama_index , 3:10pm Sep 23: , with 7K+ registered learners, 3.5K+ members in discord! 🎉📢 Huge thanks to @ShunyuYao12 @OpenAI for 2nd lecture

6

38

199

3

2

84

Shunyu Yao

@ShunyuYao12

9 months

If you ever learn a bit of computer systems or programming, you know the most intriguing and magical idea in CS is memory. Same (will be true) for AI or at least the study of autonomous agents.

1

4

76

Shunyu Yao

@ShunyuYao12

1 year

ToT achieves 10x perf by leveraging LLM's ability to 1. generate diverse choices of intermediate "thoughts" toward problem solving 2. self-evaluate thoughts via deliberate reasoning With 3. search algorithms (e.g., bfs/dfs) that help systematically explore the problem space

4

8

73

Shunyu Yao

@ShunyuYao12

6 months

When I first saw Tree of Thoughts I also asked myself this😀 great exploration into if next-token prediction can simulate search, and if you're interested in this you probably also wanna check out last paragraph

Cognitive Architectures for Language Agents

Recent efforts have augmented large language models (LLMs) with external resources (e.g., the Internet) or internal control flows (e.g., prompt chaining) for tasks requiring grounding or...

arxiv.org

noahdgoodman

@noahdgoodman

6 months

When I first saw Tree of Thoughts, I asked myself: If language models can reason better by searching, why don't they do it themselves during Chain of Thought? Some possible answers (and a new paper): 🧵

8

48

343

3

0

74

Shunyu Yao

@ShunyuYao12

1 year

New preprint time :) We propose Referral-Augmented Retrieval (RAR), an extremely simple augmentation technique that significantly improves zero-shot information retrieval. Led by awesome undergrad @_michaeltang_ , w/ @jyangballin @karthik_r_n

1

11

66

Shunyu Yao

@ShunyuYao12

1 year

If intelligence is "emergent complex behavior", then Autonomous Language Agents (ALA) like BabyAGI and AutoGPT start to enter that arena? Will revise my slides & a blogpost draft about ALA w.r.t. recent progress and share soon Quick thoughts👇 (1/n)

Siqi Chen

@blader

1 year

the top three trending repos on github are all self-prompting “primitive agi” projects: 1) babyagi by @yoheinakajima 2) autogpt by @SigGravitas 3) jarvis by @Microsoft these + scaling gets you the rest of the way there.

51

301

2K

2

11

62

Shunyu Yao

@ShunyuYao12

2 years

What if you had a bot you could just instruct in English to shop online for you? Check out our latest work 🛒WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agents w/ @__howardchen @jyangballin , @karthik_r_n @princeton_nlp

1

14

61

Shunyu Yao

@ShunyuYao12

1 year

For example, on game of 24 ("4 9 10 13"->“(13-9)*(10-4)=24”), CoT only solves 4% games --- and already fails 60% of games after generating just first 3 words! Why? LM token-by-token decoding does not allow lookahead, backtrack, or exploration of different thoughts globally.

2

3

58

Shunyu Yao

@ShunyuYao12

1 year

One of the most fundamental insights in CS: program is just a text. Two of the most amazing inventions in CS: LM and compiler. Both can turn program into some other text: reasoning or execution feedback Late night wild thoughts…

2

59

Shunyu Yao

@ShunyuYao12

1 year

Bonus 2: If you care more about "hardcore" language (instead of math/word) problems, or just enjoy CHAOS, check out our invented creative writing task! How would you find a way to chain random sentences into a coherent passage? Would you similarly plan and compare?😀

9

1

57

Shunyu Yao

@ShunyuYao12

3 months

Excited to share what I did @SierraPlatform with @noahrshinn pedram and @karthik_r_n ! 𝜏-bench evaluates critical agent capabilities omitted by current benchmarks: robustness, complex rule following, and human interaction skills. Try it out!

Karthik Narasimhan

@karthik_r_n

3 months

Excited to release 𝜏-bench (TAU for Tool-Agent-User ⚒️-🤖-🧑), a new benchmark to evaluate AI agents' performance and reliability in real-world settings with dynamic user and tool interaction. Paper: , Blog:

4

26

117

2

11

57

Shunyu Yao

@ShunyuYao12

7 months

Is it just me or haystack is too stupid for benchmarking long context capabilities? If swe-bench is too hard, at least we need tasks that require reasoning over multiple parts of long context in multiple steps

8

1

53

Shunyu Yao

@ShunyuYao12

6 months

Will visit @agihouse_org for the first time this Saturday and talk about SWE-agent, Agent-Computer Interface (ACI), and answer questions😃

John Yang

@jyangballin

6 months

SWE-agent is our new system for autonomously solving issues in GitHub repos. It gets similar accuracy to Devin on SWE-bench, takes 93 seconds on avg + it's open source! We designed a new agent-computer interface to make it easy for GPT-4 to edit+run code

64

426

2K

0

7

53

Shunyu Yao

@ShunyuYao12

5 months

@karthik_r_n @princeton_nlp @PrincetonPLI My thesis into 2 takeaways: 1. Language is the general-purpose representation for various external environments and internal thoughts. Thus, language agent is general. 2. Language reasoning can be seen as internal actions for agents. Thus, language agent is special.

0

4

52

Shunyu Yao

@ShunyuYao12

1 year

Read more about - how ToT modularizes thought decomposition/generation/evaluation & search algorithm to suit diverse tasks - formal framework (rare in prompting era) - many more experiments & findings - Inspirations from CogSci & Root of AI (eg )!

2

50

Shunyu Yao

@ShunyuYao12

2 years

Updates: - Jupyter notebooks to try out ReAct prompting with GPT-3: - 5-min video explaining ReAct: - Oral presentations at NeurIPS FMDM, EMNLP EvoNLP & NILLI workshops, happy to chat in New Orleans/Abu Dhabi and meet new friends!

GitHub - ysymyth/ReAct: [ICLR 2023] ReAct: Synergizing Reasoning and Acting in Language Models

[ICLR 2023] ReAct: Synergizing Reasoning and Acting in Language Models - ysymyth/ReAct

github.com

Shunyu Yao

@ShunyuYao12

2 years

Large Language Models (LLM) are 🔥in 2 ways: 1.🧠Reason via internal thoughts (explain jokes, math reasoning..) 2.💪Act in external worlds (SayCan, ADEPT ACT-1, WebGPT..) But so far 🧠and💪 remain distinct methods/tasks... Why not 🧠+💪? In our new work ReAct, we show 1+1>>2!

10

66

389

7

50

Shunyu Yao

@ShunyuYao12

2 months

Will we use llm as a text tool or llm use us as multimodal tool? After all we might be good at and enjoy different things😀

2

0

48

Shunyu Yao

@ShunyuYao12

3 years

Hierarchical structure is a core aspect of language syntax. Recurrent networks can systematically process recursion by emulating stacks, but can self-attention networks? If so, how? Our #ACL2021 paper shed lights into this fundamental issue! (1/5)

1

6

46

Shunyu Yao

@ShunyuYao12

6 months

in some sense, math is the first programming language, and mathematician's mind (+scratchpad) is the first compiler

3

1

45

Shunyu Yao

@ShunyuYao12

2 years

Going to NeurIPS and present my Phd papers in person for the FIRST time😀! Anyone interested in WebShop (), ReAct (), building language agents, language grounding/interaction/theory, NLP + RL...Let's DM and SCHEDULE A CHAT😆! (1/3)

2

42

Shunyu Yao

@ShunyuYao12

6 months

What we call reasoning in AI is algorithm in CS

2

3

41

Shunyu Yao

@ShunyuYao12

1 year

Had a great hour w/ @hwchase17 @charles_irl @mbusigin @yoheinakajima talking about autonomous language agents, ReAct, LangChain, BabyAGI, context management, critic, safety, and many more. Look forward to more @LangChainAI webinars, they're awesome! Replay at the same link 👇

Harrison Chase

@hwchase17

1 year

Our webinar on agents starts in 1 hour It's the most popular webinar we've hosted yet, so we had to bring in the best possible moderator: @charles_irl Come join Charles, myself, @ShunyuYao12 , @mbusigin and @yoheinakajima for some lively discussion :)

19

48

350

3

6

41

Shunyu Yao

@ShunyuYao12

6 months

SWE-agent led by amazing @jyangballin @_carlosejimenez , first authors of SWE-bench Besides code base, also check out our discord

Join the SWE-agent Discord Server!

Check out the SWE-agent community on Discord - hang out with 2173 other members and enjoy free voice and text chat.

discord.com

carlos

@_carlosejimenez

6 months

SWE-Agent is an open-source software engineering agent with a 12.3% resolve rate on SWE-Bench! Check out SWE-agent in action at Repo:

30

112

561

2

5

40

Shunyu Yao

@ShunyuYao12

1 year

Check out our new preprint led by @RTomMcCoy and thank him for getting me to know the word 'ember'🔥 Tldr: language models (LMs) are not humans, just like planes are not birds. So analyzing LMs shouldn't just use human behavior or performance tests!

Tom McCoy

@RTomMcCoy

1 year

🤖🧠NEW PAPER🧠🤖 Language models are so broadly useful that it's easy to forget what they are: next-word prediction systems Remembering this fact reveals surprising behavioral patterns: 🔥Embers of Autoregression🔥 (counterpart to "Sparks of AGI") 1/8

36

292

1K

3

2

40

Shunyu Yao

@ShunyuYao12

1 year

Thanks @USC_ISI @HJCH0 ! I talked about Formulation (CoALA) and Evaluation (Collie/InterCode/WebShop) of language agents, two directions that I find - important but understudied - academia could uniquely contribute! slides: video:

On Formulating and Evaluating Language Agents

Date Presented: 10/05/2023Speaker: Shunyu Yao, Princeton UniversityAbstract:Language agents are AI systems that use large language models (LLMs) to interact ...

www.youtube.com

Justin Cho 조현동

@HJCH0

1 year

We had the pleasure of @ShunyuYao12 give us a talk at USC ISI's NL Seminar "On Formulating and Evaluating Language Agents🤖" Check out his recorded talk to learn about a unified taxonomy for work on language agents and the next steps forward on evaluating them for complex tasks!

1

0

11

1

4

38

Shunyu Yao

@ShunyuYao12

2 years

ICLR week! Finally muster up a long-due tweet for our spotlight work: Linking Emergent and Natural Languages via Cospus Transfer paper: code: poster: Apr 27 13:30 - 15:30 EDT 1/n

2

8

39

Shunyu Yao

@ShunyuYao12

1 year

Bonus: if you're a crosswords fan, check out how ToT plays 😀 We improve game success from 1% -> 20%, but incorporating better search algorithms (e.g. how you maintain your thoughts) and heuristics (e.g. how you prune) should further enhance LLM!

1

3

37

Shunyu Yao

@ShunyuYao12

2 years

@AdeptAILabs This is super cool! We had a similar research idea in one domain (shopping), but it'd be much more powerful to train a multitask general language agent

WebShop

Towards Scalable Real-World Web Interaction with Grounded Language Agents

webshop-pnlp.github.io

1

4

37

Shunyu Yao

@ShunyuYao12

4 years

@ybisk @_jessethomason_ I find it hard too...

1

0

36

Shunyu Yao

@ShunyuYao12

4 years

Happy to announce our new #emnlp2020 paper “Keep CALM and Explore: Language Models for Action Generation in Text-based Games” is online! w/ Rohan, @mhauskn , @karthik_r_n arxiv: code: more below (1/n)

GitHub - princeton-nlp/calm-textgame: [EMNLP 2020] Keep CALM and Explore: Language Models for...

[EMNLP 2020] Keep CALM and Explore: Language Models for Action Generation in Text-based Games - princeton-nlp/calm-textgame

github.com

1

9

36

Shunyu Yao

@ShunyuYao12

1 year

Please retweet or spread the words to help the reputation of ToT, I would appreciate it! Offical ToT codebase is not

GitHub - princeton-nlp/tree-of-thought-llm: [NeurIPS 2023] Tree of Thoughts: Deliberate Problem...

[NeurIPS 2023] Tree of Thoughts: Deliberate Problem Solving with Large Language Models - princeton-nlp/tree-of-thought-llm

github.com

2

14

36

Shunyu Yao

@ShunyuYao12

4 years

For autonomous tasks with language (e.g. text games), how much does an agent rely on language semantics vs. memorization? Our #NAACL2021 paper (, joint w/ @karthik_r_n , @mhauskn ) proposes ablation studies with surprising findings and useful insights! (1/3)

1

2

32

Shunyu Yao

@ShunyuYao12

2 years

Flying to #EMNLP for the first #NLP conference in my life, despite being an OLD fourth year phd student😂😂 Would love to meet new friends and chat about language grounding and interaction!

2

0

32

Shunyu Yao

@ShunyuYao12

6 months

Attention is all we need. Memory is all we have.

0

2

31

Shunyu Yao

@ShunyuYao12

9 months

Just realize another analog of humans and LLMs: we develop and evaluate them on what is easy to evaluate (SAT or MMLU) then set them out on what is hard to evaluate.

1

5

30

Shunyu Yao

@ShunyuYao12

1 year

@noahshinn024 et al did Reflexion in Mar 2023, and tons of LLM-critic projects since. Still, we worked on Reflexion v2 . What for? - clean & general conceptual framework via language agent/RL - strong empirical perf on more diverse & complex tasks (1/n)

2

28

Shunyu Yao

@ShunyuYao12

11 months

Agent is not just future but now

0

3

27

Shunyu Yao

@ShunyuYao12

2 years

1. ReAct > 🧠/💪only methods, e.g. - On knowledge reasoning tasks, interacting with wiki API obtains new knowledge and avoids hallucination. - On decision making tasks, sparse+flexible thoughts can decompose goal, plan actions, induce commonsense, track progress, adjust plan..

1

27

Shunyu Yao

@ShunyuYao12

7 months

This meme will fly high as swe agents fly high

Shunyu Yao

@ShunyuYao12

1 year

Meme aside, Check SWE-bench that hits many checks for a good benchmark - hard but useful to solve, easy to evaluate - automatically constructed from real GitHub issues and pull requests - challenge super long context, retrieval, coding, etc - can easily update with new instances

1

10

88

0

1

27

Shunyu Yao

@ShunyuYao12

10 months

Can someone tell me whats qstar

9

1

26

Shunyu Yao

@ShunyuYao12

1 year

Tree of Thoughts is a serious paper and serious research, not a github star chasing leverage. I appreciate any implementation of any of my work, but it should link to what is the offical implementation to avoid confusion and abuse.

2

26

Shunyu Yao

@ShunyuYao12

7 months

@cognition_labs we will release some interesting research results on coding agents soon😀

0

26

Shunyu Yao

@ShunyuYao12

5 months

Congrats @andrewwhite01 and team! Still one of the coolest applications of ReAct and language agents, and one of my personal favorite!

Andrew White 🐦‍⬛

@andrewwhite01

5 months

ChemCrow is out today in @NatMachIntell ! ChemCrow is an agent that uses chem tools and a cloud-based robotic lab for open-ended chem tasks. It’s been a journey to get to publication and I’d like to share some history about it. It started back in 2022. 1/8

16

74

460

0

2

25

Shunyu Yao

@ShunyuYao12

1 year

Great blogpost about recent advances in Autonomous Language Agents -- now can try them all in LangChain

Harrison Chase

@hwchase17

1 year

🤖Autonomous Agents & Agent Simulations🤖 Four agent-related projects (AutoGPT, BabyAGI, CAMEL, and Generative Agents) have exploded recently We wrote a blog on they differ from previous @LangChainAI agents and how we've incorporated some key ideas

31

168

828

0

4

25

Shunyu Yao

@ShunyuYao12

10 months

Cool to see followup efforts using Tree of Thoughts () for important applications (LLM safety and jailbreaking) ... And a growth of Tree-of-x work😂

Tree of Thoughts: Deliberate Problem Solving with Large Language Models

Language models are increasingly being deployed for general problem solving across a wide range of tasks, but are still confined to token-level, left-to-right decision-making processes during...

arxiv.org

Amin Karbasi

@aminkarbasi

10 months

In collaboration with @robusthq , yesterday we shared "Tree of Attacks" a method than can jailbreak @OpenAI GPT-4 like 90% of the times. It was just covered in @wired

4

50

236

1

0

25

Shunyu Yao

@ShunyuYao12

1 year

Excited about this work on emergent communication (EC)! EC's been a tricky subject (i.e. lots of toy papers), but IMO the true potential is unleashing soon. Simplest reason: we're running out of language by humans on Internet. Have to use machine's self-generated language soon!

Yao Mark Mu

@YaoMarkMu1

1 year

Our new paper, EC^2, has been published in CVPR2023. It presents a novel video-language per-training scheme via emergent communication for few-shot embodied control. Project page: Paper:

0

8

39

2

24

Shunyu Yao

@ShunyuYao12

1 year

@DrJimFan 's "no-gradient architecture" is exactly what we call "verbal reinforcement learning". Awesome progress in this direction using a great testbed! It is fair to say we (significantly) haven't reached the capabilitity limit out of just calling gpt4 apis. Still much to do!

Jim Fan

@DrJimFan

1 year

What if we set GPT-4 free in Minecraft? ⛏️ I’m excited to announce Voyager, the first lifelong learning agent that plays Minecraft purely in-context. Voyager continuously improves itself by writing, refining, committing, and retrieving *code* from a skill library. GPT-4 unlocks

365

2K

9K

3

2

23

Shunyu Yao

@ShunyuYao12

9 months

Never done pair programming but did pair prompting for the first time today😂

0

22

Shunyu Yao

@ShunyuYao12

2 years

3. ReAct naturally produces more interpretable and trustworthy trajs, where humans can - inspect fact source (internal vs. external) - check reasoning basis of decisions - modify model thoughts for policy edit on-the-go, an exciting new paradigm for human-machine interaction!

1

20

Shunyu Yao

@ShunyuYao12

2 years

2. ReAct generalizes strongly, both in few-shot prompting and finetuning. e.g. - On WebShop/AlfWorld, 1/2-shot ReAct outperforms imitation learning w/ 3k/100k samples by 10/34%! - Using LLM ReAct trajs, finetuned smaller LMs outperforms LLM and finetuned🧠/💪only models!

1

19

Shunyu Yao

@ShunyuYao12

7 months

Tom is AWESOME to work with, from most high-level ideas to most low-level details. Being his postdoc would be great fun and growing experience!

Tom McCoy

@RTomMcCoy

7 months

I am hoping to hire a postdoc who would start in Fall 2024. If you are interested in the intersection of linguistics, cognitive science, and AI, I encourage you to apply! Please see this link for details:

2

53

164

1

19

Shunyu Yao

@ShunyuYao12

5 months

misalignment of ai and humans is not as dangerous as misalignment of our own behavior and desire, or misalignment among our various desires, or misalignment among various us

1

17

Shunyu Yao

@ShunyuYao12

1 year

from @princeton_nlp @DeepMind @cocosci_lab

3

0

19

Shunyu Yao

@ShunyuYao12

6 months

@AndrewYNg

Cognitive Architectures for Language Agents

Recent efforts have augmented large language models (LLMs) with external resources (e.g., the Internet) or internal control flows (e.g., prompt chaining) for tasks requiring grounding or...

arxiv.org

1

0

19

Shunyu Yao

@ShunyuYao12

6 months

Guys, before achieving AGI, we need to solve zork-1?

2

19

Shunyu Yao

@ShunyuYao12

2 years

ReAct = Synergize [Rea]soning and [Act]ing in LM How? ReAct prompts LLM with human task-solving trajectories with interleaving 🧠flexible thoughts and 💪domain-specific actions, so that it can generate both. Why? Strong generalization on VERY diverse tasks + ALIGNMENT benefits!

1

0

19

Shunyu Yao

@ShunyuYao12

2 years

As GPT context length keeps increasing, will all retrieval become in-context retrieval? Or is (traditional) retrieval the key to increasing context length...

2

18

Shunyu Yao

@ShunyuYao12

9 months

Anyone with updated research belief after #NeurIPS2023 week?

2

1

18

Shunyu Yao

@ShunyuYao12

1 year

Very cool work, analyzing the risks and robustness of ReAct agents across scenarios and base LLMs! This direction will be very important.

Yangjun Ruan

@YangjunR

1 year

Should you let LMs control your email? terminal? bank account? or even your smart home?🤔 🔥Introducing ToolEmu for identifying risks associated with LM agents at scale! 🛠️Featuring LM-emulation of tools & automated realistic risk detection 🚨GPT4 is risky in 40% of our cases!

6

48

173

0

4

17

Shunyu Yao

@ShunyuYao12

10 months

@DrJimFan @RichardSSutton Ai will develop new concepts and symbols and tranfer back to humans.

Linking Emergent and Natural Languages via Corpus Transfer

The study of language emergence aims to understand how human languages are shaped by perceptual grounding and communicative intent. Computational approaches to emergent communication (EC)...

arxiv.org

1

17

Shunyu Yao

@ShunyuYao12

9 months

If #GPT4 is open sourced tmr, what would/could you even do with it? How is that gonna change AI?

7

0

17

Shunyu Yao

@ShunyuYao12

1 year

It's crazy most web agent work in academia is still doing MiniWoB...

1

0

17

Shunyu Yao

@ShunyuYao12

2 years

I guess the success of diffusion in vision and chain of thought in language shares something in common: solving things step by step is just good

2

1

17

Shunyu Yao

@ShunyuYao12

1 year

@srush_nlp 🤣i would say higher but not exponential, given search has heuristics (eg bfs prune breadth, dfs prune subtree). But hopefully we can (and should) use open and free models soon

1

17

Shunyu Yao

@ShunyuYao12

11 months

Is the prompt engineering mess after GPT-3 till now similar to "network architecture/loss engineering" mess after AlexNet till now?

2

0

17

Shunyu Yao

@ShunyuYao12

1 year

I opened an issue at just to ask the repo links to our offical repo to avoid any confusion, but closed by @KyeGomezB without any resolving. I don't like it.

link to the official implementation: https://github.com/ysymyth/tree-of-thought-llm · Issue #54 ·...

hi @kyegomez thanks for implementing our work! as pointed out in princeton-nlp/tree-of-thought-llm#17, would you mind linking to our official repo in your README.md to avoid any confusion? thanks i...

github.com

3

16

Shunyu Yao

@ShunyuYao12

8 months

Cool work and great extension of WebShop () --- We've been thinking about making it personalized for some years, it's great someone starts doing it!

WebShop: Towards Scalable Real-World Web Interaction with Grounded...

Existing benchmarks for grounding language in interactive environments either lack real-world linguistic elements, or prove difficult to scale up due to substantial human involvement in the...

arxiv.org

Zonghan Yang

@yang_zonghan

8 months

We all know that __alignment__ elicits the capability of foundation models (FMs), while __agents__ autonomize FMs as copilots. Well, I'd say the two are intrinsically intertwined! Excited to introduce the principles of unified alignment for agents: (1/N)

5

11

62

2

15

Shunyu Yao

@ShunyuYao12

7 months

Is the timing coincidence or planned? I think planned😂

varepsilon

@var_epsilon

8 months

mogged again

78

1K

14K

2

0

16

Shunyu Yao

@ShunyuYao12

1 year

@kushirosea Our work explores simpler search algos like bfs/dfs, but mcts is def a natural todo!

2

0

16

Shunyu Yao

@ShunyuYao12

10 months

When is language model enough and when is language agent needed? Seems a good probe is whether humans speak or write. Eg. We can answer most questions fluently in everyday conversations, but need iterative revisions for writing math proof, blogposts, code, etc.

1

0

16

Shunyu Yao

@ShunyuYao12

2 years

Check the paper for MUCH MORE findings and insights! With @jezhao @Dian_Yu0 Nan Izhak @karthikn @caoyuan33 @googleai @princetonnlp

2

1

15

Shunyu Yao

@ShunyuYao12

1 year

Going to #ACL2023NLP next week (10-12), lmk if you want to chat😀

0

15

Shunyu Yao

@ShunyuYao12

1 year

I also first saw Langchain when it implemented ReAct at 0.0.3 with <100 stars. Now it's 0.0.131 with 20k+ stars. A lot of hard work! Great demos every day (a lot with super easy zero-shot-react-agent!). Congrats @hwchase17 and @LangChainAI and look forward to the future!

Miles Grimshaw

@milesgrimshaw

1 year

I first saw Langchain on Twitter when Harrison implemented the ReAct paper, exposing the LLM’s reasoning: . I was impressed w/ the elegant abstractions he wrote & captivated by the possibilities of orchestrating LLMs as an intelligence layer.

1

0

6

0

15