Bibliography:

1. ​ NN Inner Monologue

2. ​ ‘GPT’ tag

3. ​ ‘NN sampling’ tag

4. ​ ‘GPT-4’ tag

5. ​ ‘instruct-tuning LLMs’ tag

6. ​ ‘PaLM’ tag

7. ​ ‘inner-monologue (psych)’ tag

8. ​ ‘meta-learning’ tag

9. ​ Free-Play Periods for RL Agents

10. ​ It Looks Like You’re Trying To Take Over The World

11. ​ O1 Turns Pro

12. ​ Procedural Knowledge in Pretraining Drives Reasoning in Large Language Models

13. ​ Mind Your Step (by Step): Chain-of-Thought can Reduce Performance on Tasks where Thinking Makes Humans Worse

14. ​ Thinking LLMs: General Instruction Following with Thought Generation

15. ​ When a language model is optimized for reasoning, does it still show embers of autoregression? An analysis of OpenAI o1

16. ​ Evaluation of OpenAI o1: Opportunities and Challenges of AGI

17. ​ LLMs Still Can’t Plan; Can LRMs? A Preliminary Evaluation of OpenAI’s o1 on PlanBench

18. ​ Training Language Models to Self-Correct via Reinforcement Learning

19. ​ To CoT or not to CoT? Chain-of-thought helps mainly on math and symbolic reasoning

20. ​ Physics of Language Models: Part 2.1, Grade-School Math and the Hidden Reasoning Process

21. ​ Connecting the Dots: LLMs can Infer and Verbalize Latent Structure from Disparate Training Data

22. ​ Can Long-Context Language Models Subsume Retrieval, RAG, SQL, and More?

23. ​ OlympicArena: Benchmarking Multi-discipline Cognitive Reasoning for Superintelligent AI

24. ​ How Far Can Transformers Reason? The Locality Barrier and Inductive Scratchpad

25. ​ OmegaPRM: Improve Mathematical Reasoning in Language Models by Automated Process Supervision

26. ​ MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark

27. ​ A Theoretical Understanding of Self-Correction through In-context Alignment

28. ​ Intelligent Go-Explore (IGE): Standing on the Shoulders of Giant Foundation Models

29. ​ From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Step

30. ​ Retrieval Head Mechanistically Explains Long-Context Factuality

31. ​ Let’s Think Dot by Dot: Hidden Computation in Transformer Language Models

32. ​ Autonomous LLM-driven research from data to human-verifiable research papers

33. ​ Missed Connections: Lateral Thinking Puzzles for Large Language Models

34. ​ ChatGPT Can Predict the Future when it Tells Stories Set in the Future About the Past

35. ​ Visualization-of-Thought Elicits Spatial Reasoning in Large Language Models

36. ​ Do language models plan ahead for future tokens?

37. ​ FABLES: Evaluating faithfulness and content selection in book-length summarization

38. ​ Re-evaluating GPT-4’s bar exam performance

39. ​ Long-form factuality in large language models

40. ​ Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking

41. ​ RNNs are not Transformers (Yet): The Key Bottleneck on In-context Retrieval

42. ​ Tokenization counts: the impact of tokenization on arithmetic in frontier LLMs

43. ​ Chain-of-Thought Empowers Transformers to Solve Inherently Serial Problems

44. ​ Same Task, More Tokens: the Impact of Input Length on the Reasoning Performance of Large Language Models

45. ​ Why are Sensitive Functions Hard for Transformers?

46. ​ Chain-of-Thought Reasoning Without Prompting

47. ​ V-STaR: Training Verifiers for Self-Taught Reasoners

48. ​ More Agents Is All You Need

49. ​ The Impact of Reasoning Step Length on Large Language Models

50. ​ Large Language Models Play StarCraft II: Benchmarks and A Chain of Summarization Approach

51. ​ Beyond Human Data: Scaling Self-Training for Problem-Solving with Language Models (ReST^EM)

52. ​ Tree of Attacks (TAP): Jailbreaking Black-Box LLMs Automatically

53. ​ Universal Self-Consistency for Large Language Model Generation

54. ​ Can Generalist Foundation Models Outcompete Special-Purpose Tuning? Case Study in Medicine

55. ​ Training Chain-of-Thought via Latent-Variable Inference

56. ​ Compositional Capabilities of Autoregressive Transformers: A Study on Synthetic, Interpretable Tasks

57. ​ On Measuring Faithfulness or Self-consistency of Natural Language Explanations

58. ​ Zero-Shot Goal-Directed Dialogue via RL on Imagined Conversations

59. ​ Large Language Models can Strategically Deceive their Users when Put Under Pressure

60. ​ Rephrase and Respond: Let Large Language Models Ask Better Questions for Themselves

61. ​ Everything of Thoughts: Defying the Law of Penrose Triangle for Thought Generation

62. ​ Implicit Chain-of-Thought Reasoning via Knowledge Distillation

63. ​ Preventing Language Models From Hiding Their Reasoning

64. ​ Branch-Solve-Merge Improves Large Language Model Evaluation and Generation

65. ​ Can GPT models be Financial Analysts? An Evaluation of ChatGPT and GPT-4 on mock CFA Exams

66. ​ The Expressive Power of Transformers with Chain-of-Thought

67. ​ Language Agent Tree Search Unifies Reasoning Acting and Planning in Language Models

68. ​ Large Language Models Cannot Self-Correct Reasoning Yet

69. ​ Think before you speak: Training Language Models With Pause Tokens

70. ​ Embers of Autoregression: Understanding Large Language Models Through the Problem They are Trained to Solve

71. ​ Contrastive Decoding Improves Reasoning in Large Language Models

72. ​ Re-Reading Improves Reasoning in Large Language Models

73. ​ From Sparse to Dense: GPT-4 Summarization with Chain of Density (CoD) Prompting

74. ​ Graph of Thoughts: Solving Elaborate Problems with Large Language Models

75. ​ Solving Challenging Math Word Problems Using GPT-4 Code Interpreter with Code-based Self-Verification

76. ​ Android in the Wild: A Large-Scale Dataset for Android Device Control

77. ​ LLMs as Workers in Human-Computational Algorithms? Replicating Crowdsourcing Pipelines with LLMs

78. ​ TableGPT: Towards Unifying Tables, Nature Language and Commands into One GPT

79. ​ Question Decomposition Improves the Faithfulness of Model-Generated Reasoning

80. ​ Measuring Faithfulness in Chain-of-Thought Reasoning

81. ​ Unleashing the Emergent Cognitive Synergy in Large Language Models: A Task-Solving Agent through Multi-Persona Self-Collaboration

82. ​ Explaining Competitive-Level Programming Solutions using LLMs

83. ​ Teaching Arithmetic to Small Transformers

84. ​ Language models are weak learners

85. ​ Let’s Do a Thought Experiment: Using Counterfactuals to Improve Moral Reasoning

86. ​ GKD: Generalized Knowledge Distillation for Auto-regressive Sequence Models

87. ​ Large Language Models as Tax Attorneys: A Case Study in Legal Capabilities Emergence

88. ​ Iterative Translation Refinement with Large Language Models

89. ​ Thought Cloning: Learning to Think while Acting by Imitating Human Thinking

90. ​ Let’s Verify Step by Step

91. ​ Towards Revealing the Mystery behind Chain-of-Thought: A Theoretical Perspective

92. ​ Improving Factuality and Reasoning in Language Models through Multiagent Debate

93. ​ How Language Model Hallucinations Can Snowball

94. ​ Tree of Thoughts (ToT): Deliberate Problem Solving with Large Language Models

95. ​ Large Language Model Programs

96. ​ Language Models Don’t Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting

97. ​ Distilling Step-by-Step! Outperforming Larger Language Models with Less Training Data and Smaller Model Sizes

98. ​ Decomposition Enhances Reasoning via Self-Evaluation Guided Decoding

99. ​ LLM+P: Empowering Large Language Models with Optimal Planning Proficiency

100. ​ Boosting Theory-of-Mind Performance in Large Language Models via Prompting

101. ​ Think Before You Act: Unified Policy for Interleaving Language Reasoning with Actions

102. ​ Language Models can Solve Computer Tasks

103. ​ Reflexion: Language Agents with Verbal Reinforcement Learning

104. ​ How well do Large Language Models perform in Arithmetic tasks?

105. ​ SelfCheckGPT: Zero-Resource Black-Box Hallucination Detection for Generative Large Language Models

106. ​ Language Is Not All You Need: Aligning Perception with Language Models (Kosmos-1)

107. ​ Multimodal Chain-of-Thought Reasoning in Language Models

108. ​ Faithful Chain-of-Thought Reasoning

109. ​ Large Language Models are Versatile Decomposers: Decompose Evidence and Questions for Table-based Reasoning

110. ​ ChatGPT Goes to Law School

111. ​ Large Language Models as Fiduciaries: A Case Study Toward Robustly Communicating With Artificial Intelligence Through Legal Standards

112. ​ Interactive-Chain-Prompting (INTERCPT): Ambiguity Resolution for Crosslingual Conditional Generation with Interaction

113. ​ Iterated Decomposition: Improving Science Q&A by Supervising Reasoning Processes

114. ​ Solving math word problems with process & outcome-based feedback

115. ​ PAL: Program-aided Language Models

116. ​ Measuring Progress on Scalable Oversight for Large Language Models

117. ​ U-PaLM: Transcending Scaling Laws with 0.1% Extra Compute

118. ​ Large Language Models Can Self-Improve

119. ​ Challenging BIG-Bench Tasks (BBH) and Whether Chain-of-Thought Can Solve Them

120. ​ Self-Ask: Measuring and Narrowing the Compositionality Gap in Language Models (Bamboogle)

121. ​ Language Models are Multilingual Chain-of-Thought Reasoners

122. ​ ReAct: Synergizing Reasoning and Acting in Language Models

123. ​ Dynamic Prompt Learning via Policy Gradient for Semi-structured Mathematical Reasoning

124. ​ FOLIO: Natural Language Reasoning with First-Order Logic

125. ​ Faithful Reasoning Using Large Language Models

126. ​ Limitations of Language Models in Arithmetic and Symbolic Induction

127. ​ Language Models Can Teach Themselves to Program Better

128. ​ Language Model Cascades

129. ​ CodeT: Code Generation with Generated Tests

130. ​ Can large language models reason about medical questions?

131. ​ Inner Monologue: Embodied Reasoning through Planning with Language Models

132. ​ Exploring Length Generalization in Large Language Models

133. ​ Language Models (Mostly) Know What They Know

134. ​ Solving Quantitative Reasoning Problems with Language Models

135. ​ Maieutic Prompting: Logically Consistent Reasoning with Recursive Explanations

136. ​ Large Language Models are Zero-Shot Reasoners

137. ​ Instruction Induction: From Few Examples to Natural Language Task Descriptions

138. ​ Least-to-Most Prompting Enables Complex Reasoning in Large Language Models

139. ​ Dialog Inpainting: Turning Documents into Dialogues

140. ​ Unifying Language Learning Paradigms

141. ​ Can language models learn from explanations in context?

142. ​ Socratic Models: Composing Zero-Shot Multimodal Reasoning with Language

143. ​ STaR: Bootstrapping Reasoning With Reasoning

144. ​ A Conversational Paradigm for Program Synthesis

145. ​ Self-Consistency Improves Chain-of-Thought Reasoning in Language Models

146. ​ Learning-by-Narrating: Narrative Pre-Training for Zero-Shot Dialogue Comprehension

147. ​ PromptChainer: Chaining Large Language Model Prompts through Visual Programming

148. ​ Chain-of-Thought Prompting Elicits Reasoning in Large Language Models

149. ​ Reasoning Like Program Executors

150. ​ A Neural Network Solves and Generates Mathematics Problems by Program Synthesis: Calculus, Differential Equations, Linear Algebra, and More

151. ​ DREAM: Uncovering Mental Models behind Language Models

152. ​ Reframing Human-AI Collaboration for Generating Free-Text Explanations

153. ​ NeuroLogic A^✱esque Decoding: Constrained Text Generation with Lookahead Heuristics

154. ​ WebGPT: Improving the factual accuracy of language models through web browsing

155. ​ Few-Shot Self-Rationalization with Natural Language Prompts

156. ​ Training Verifiers to Solve Math Word Problems

157. ​ Unsupervised Neural Machine Translation with Generative Language Models Only

158. ​ Show Your Work: Scratchpads for Intermediate Computation with Language Models

159. ​ AI Chains: Transparent and Controllable Human-AI Interaction by Chaining Large Language Model Prompts

160. ​ Teaching Autoregressive Language Models Complex Tasks By Demonstration

161. ​ Program Synthesis with Large Language Models

162. ​ Decision Transformer: Reinforcement Learning via Sequence Modeling

163. ​ Explainable Multi-hop Verbal Reasoning Through Internal Monologue

164. ​ A simple method to keep GPT-3 focused in a conversation

165. ​ Measuring Mathematical Problem Solving With the MATH Dataset

166. ​ Prompt Programming for Large Language Models: Beyond the Few-Shot Paradigm

167. ​ How We Accidentally Gave our Bots Their Personalities

168. ​ Word in Context: Agent and Agent Clarification (69% Dev)

169. ​ I found that getting GPT-3 to add its own "internal monologue" in parentheses to be a helpful strategy…

170. ​ Seems to work

171. ​ Teaching GPT-3 to do a brute force 'for loop' checking answers also seems to work

172. ​ Inducing Self-Explanation: a Meta-Analysis

173. ​ Program Induction by Rationale Generation: Learning to Solve and Explain Algebraic Word Problems

174. ​ Why Do Humans Reason? Arguments for an Argumentative Theory

175. ​ ba03e8d7db678948a7585a947ea8a4eac13d6abf.pdf

176. ​ How to Dramatically Improve the Reasoning Ability of GPT-3

177. ​ 642b641a22ab789da5eba95379dfeb1e7c7596e9.html

178. ​ A Preliminary Exploration into Factored Cognition With Language Models

179. ​ WiC_SelfContextStuffingImproved_Last10_stuft_examplesNV.ipynb

180. ​ Vincent-163/transformer-Arithmetic

181. ​ Magic ToDo List Creator

182. ​ 99da0d8d421922b8768e9f6e35207d59db5bb214.html

183. ​ Short Story on AI: ‘Forward Pass’

184. ​ AI Dungeon Players Can Now Translate Their Stories into Emojis by Just Clicking a Button.

185. ​ d4a1d75abef33a907533b26f731c3ebb3ac090a1.html

186. ​ Solving Math Word Problems: We’ve Trained a System That Solves Grade School Math Problems With Nearly Twice the Accuracy of a Fine-Tuned GPT-3 Model. It Solves about 90% As Many Problems As Real Kids: a Small Sample of 9-12 Year Olds Scored 60% on a Test from Our Dataset, While Our System Scored 55% on Those Same Problems. This Is Important Because Today’s AI Is Still Quite Weak at Commonsense Multistep Reasoning, Which Is Easy Even for Grade School Kids. We Achieved These Results by Training Our Model to Recognize Its Mistakes, so That It Can Try Repeatedly Until It Finds a Solution That Works

187. ​ Prompting Diverse Ideas: Increasing AI Idea Variance

188. ​ Teaching a Neural Network to Use a Calculator

189. ​ 3a2a05d3c82f879dc7abd75a01d242a08925804f.html

190. ​ Connecting the Dots: LLMs Can Infer & Verbalize Latent Structure from Training Data

191. ​ Preventing Language Models from Hiding Their Reasoning

192. ​ Steganography in Chain-Of-Thought Reasoning

193. ​ Visible Thoughts Project and Bounty Announcement

194. ​ I Think ‘GPT-3 Can’t Do Parity Checking’ Isn’t Quite Right. It Can Clearly Pattern Match the Algorithm, Almost Perfectly. It’s Just a Little Mistake Prone. Here, I Invented a Syntax for Having It Evaluate Parity on Each Pair of Digits. It...almost Gets It Right.

195. ​ design#future-tag-features

     [Transclude the forward-link's context]

196. ​ 2023-chen-table1-gpt35promptsusedtorepeatedlyrefinenaturallanguagetranslationsinnermonologuestyle.png

197. ​ 2023-lee-figure1-numberformattingforgpt2arithmetic.jpg

198. ​ 2023-lee-figure2-thefourinputformattingoptionsforgptinnermonologue.png

199. ​ 2023-lee-figure3-performanceofgpton3digitarithmeticdependsondatadistribution.png

200. ​ 2023-lee-figure9-arithmeticcanbelearnedevenwithnoiseintheinnermonologuetranscripts.jpg

201. ​ 2023-moghaddam-figure1-examplesofzerovstwoshottheoryofmindprompting.png

202. ​ 2023-moghaddam-figure3-gpt3andgpt4performanceontheoryofmindwithinnermonologues.jpg

203. ​ 2023-pilaut-figure2-exampleambiguitiesintranslatingfrenchtoenglish.jpg

204. ​ 2023-pilaut-figure3-interceptinnermonologuequestionaskingonlyemergesatscalefrompalm62bto540b.png

205. ​ 2023-lee-figure6-sampleefficiencyofvariousinnermonologueformatsshowingmoredetailedisbetterforimitationlearning.png

206. ​ 2022-10-24-raldi-gpt3doesanastonishinglygoodjobcreatingbothsidesofaninteractivefictiontranscript.html

207. ​ 2022-05-28-gpt3user-thinkingisallyouneed.html

208. ​ 2022-dai-figure4-qreccretrevialperformancelogscalinginwikidialogdatasetsize.jpg

209. ​ 2022-huang-figure2-3kindsofnaturallanguagefeedbackforcontrollingsaycaninnermonologue.png

210. ​ 2022-huang-figure3-testinginnermonologuein3roboticdomains.png

211. ​ 2022-huang-figure5a-emergentcapabilities-continuedadaptationtonewinstructions.png

212. ​ 2022-huang-figure5b-emergentcapabilities-selfproposingnewgoalsunderinfeasibilityofoldgoals.png

213. ​ 2022-huang-figure5c-emergentcapabilities-multilingualinteractioninchinese.png

214. ​ 2022-huang-figure5d-emergentcapabilities-interactivesceneunderstandinglikeshrdlu.png

215. ​ 2022-lampinen-figure2-gopherperformanceimprovementsfromexplanationofproblems.jpg

216. ​ 2022-lampinen-figure4-largermodelsbenefitmorefromexplanationofproblems.png

217. ​ 2022-press-figure3-gpt3selfaskinnermonologuedemonstration.png

218. ​ 2022-press-figure4-selfaskinnermonologueperformsequallywellon1hopand2hopquestionanswering.png

219. ​ 2022-press-figure5-selfaskplusgooglesearchengine-innermonologueforsearchingtheinternettoanswermultihopquestions.png

220. ​ 2022-press-table1-selfaskplusgooglesearchengine-innermonologueforsearchingtheinternettoanswermultihopquestions-benchmarkperformance.jpg

221. ​ 2022-shi-figure4-multilingualinnermonologuescalingbyparametercountingpt3andpalm.png

222. ​ 2022-shi-figure5-multiglinalfewshotscalinginpalm540bbynumberofexamples.png

223. ​ 2022-tay-ul2-innermonologueresults.png

224. ​ 2022-wang-figure2-selfconsistencycompletiongreatlyimprovesanswercorrectness.jpg

225. ​ 2022-wei-figure2-lamdamathwordproblemscalinginmodelparametersize.jpg

226. ​ 2022-wei-figure3-lamdamathwordproblemscalingwithmodelparametersizewhenusinginnermonologueprompts.jpg

227. ​ 2022-wei-figure5-lamdamatsymbolicreasoningproblemscalingwithmodelparametersizewhenusinginnermonologueprompts.png

228. ​ 2022-wei-figure6-lamdacommonsensereasoningproblemscalingwithmodelparametersizewhenusinginnermonologueprompts.png

229. ​ 2022-wei-figure8-lamdavsgpt3.png

230. ​ 2022-zeng-figure2-socraticmodelsworkflowoverview.png

231. ​ https://applied-llms.org/

232. ​ https://blog.valentin.sh/chatgpt5/

233. ​ a969ecc7563621c6816a28e2029162023050ce24.html

234. ​ https://builtin.com/job/customer-success/expert-ai-teacher-contract/1267315

235. ​ https://generative.ink/posts/methods-of-prompt-programming/#serializing-reasoning

236. ​ 2fa0cae05f887923f2a169fecaa094fa3075f6ba.html#serializing-reasoning

237. ​ https://github.com/OpenBioLink/ThoughtSource

238. ​ https://github.com/desik1998/MathWithLLMs

239. ​ https://github.com/ggerganov/llama.cpp/pull/1773

240. ​ https://github.com/openai/openai-cookbook/blob/main/techniques_to_improve_reliability.md

241. ​ https://jxnl.github.io/instructor/blog/2023/11/05/chain-of-density/

242. ​ b272f657beb41f780daeefd4da1c58566a855360.html

243. ​ https://lingo.csail.mit.edu/blog/arithmetic_gpt3/

244. ​ 116d0377b5d9ec8e00865c1507909dd879d619c2.html

245. ​ https://model-checking.github.io/kani-verifier-blog/2023/05/01/writing-code-with-chatgpt-improve-it-with-kani.html

246. ​ https://niplav.site/decompose.html#Small_Experiment

247. ​ https://openai.com/index/introducing-openai-o1-preview/

248. ​ https://platform.openai.com/docs/guides/reasoning/how-reasoning-works

249. ​ https://reasoning-tokens.ghost.io/reasoning-tokens/

250. ​ https://research.google/blog/google-research-2022-beyond-language-vision-and-generative-models/

251. ​ https://research.google/blog/minerva-solving-quantitative-reasoning-problems-with-language-models/

252. ​ https://statmodeling.stat.columbia.edu/2023/08/30/chatgpt-4-can-do-3-digit-multiplication/

253. ​ https://towardsdatascience.com/1-1-3-wait-no-1-1-2-how-to-have-gpt-sanity-check-itself-136e846987bf

254. ​ 505a0878a1c3e6053fee5a93169d7ef3334e7cde.html

255. ​ https://www.fhi.ox.ac.uk/wp-content/uploads/2021/08/QNRs_FHI-TR-2021-3.0.pdf

256. ​ https://www.lesswrong.com/posts/XaKLjyDejtXDoRAzL/a-quick-experiment-on-lms-inductive-biases-in-performing

257. ​ 734e0042214ff6c758d48c13b5bbeee104a17936.html

258. ​ https://www.lesswrong.com/posts/bwyKCQD7PFWKhELMr/by-default-gpts-think-in-plain-sight#zfzHshctWZYo8JkLe

259. ​ https://www.oreilly.com/radar/what-we-learned-from-a-year-of-building-with-llms-part-i/

260. ​ https://www.patterns.app/blog/2023/01/18/crunchbot-sql-analyst-gpt/

261. ​ d7aaf7b7491492af22c98dae1079fbfa93961b5b.html

262. ​ https://www.pnas.org/doi/full/10.1073/pnas.2317967121

263. ​ https://www.reddit.com/r/ChatGPT/comments/10zavbv/extending_chatgpt_with_some_additional_internal/

264. ​ 3e3c992ee499fa4294ffcb3c05421fdc5117185f.html

265. ​ https://www.reddit.com/r/ChatGPT/comments/11anct1/its_easy_to_give_chatgpt_a_bonafide_consciousness/

266. ​ 17d0d9d4c3e13ca5537f0452222b8889206c71be.html

267. ​ https://www.reddit.com/r/LocalLLaMA/comments/1fuxw8d/just_for_kicks_i_looked_at_the_newly_released/

268. ​ https://www.reddit.com/r/OpenAI/comments/1fxa6d6/two_purported_instances_of_o1preview_and_o1mini/

269. ​ https://www.reddit.com/r/OpenAI/comments/1gjj430/o1_preview_got_weird_today/

270. ​ https://www.reddit.com/r/PromptEngineering/comments/1fj6h13/hallucinations_in_o1preview_reasoning/

271. ​ https://www.reddit.com/r/slatestarcodex/comments/1201v68/10word_quote_a_short_and_simple_failure_mode_of/jdigzkh/?context=3

272. ​ https://www.waluigipurple.com/post/revising-poetry-with-gpt-4

273. ​ 5c3ed833b2e637e96256b129d4c61794163980eb.html

274. ​ https://www.youtube.com/watch?v=g7YJIpkk7KM?t=38

275. ​ https://x.com/AISafetyMemes/status/1841891795782775221

276. ​ https://x.com/BlinkDL_AI/status/1677593798531223552

277. ​ https://x.com/D_Rod_Tweets/status/1628449917898264576

278. ​ https://x.com/DaveMonlander/status/1612802240582135809

279. ​ https://x.com/KevinAFischer/status/1646018246225846272

280. ​ https://x.com/KevinAFischer/status/1646677902833102849

281. ​ https://x.com/KevinAFischer/status/1646690838981005312

282. ​ https://x.com/Kyrannio/status/1793874431179460911

283. ​ https://x.com/MParakhin/status/1632087709060825088

284. ​ https://x.com/MikePFrank/status/1622202768743096320

285. ​ https://x.com/MikePFrank/status/1622495004810784768

286. ​ https://x.com/StudentInfosec/status/1640360234882310145

287. ​ https://x.com/adiwyner/status/1629980541716922369

288. ​ https://x.com/amasad/status/1628546489843863555

289. ​ https://x.com/andrewwhite01/status/1616933106786738176

290. ​ https://x.com/deepfates/status/1682110624271319040

291. ​ https://x.com/denny_zhou/status/1547662872511070212

292. ​ https://x.com/denny_zhou/status/1587115933293678592

293. ​ https://x.com/emollick/status/1705422957856604503

294. ​ https://x.com/finereli/status/1782611247709786145

295. ​ https://x.com/gfodor/status/1626270272314839041

296. ​ https://x.com/goodside/status/1563191853587271681

297. ​ https://x.com/goodside/status/1568375796904886274

298. ​ https://x.com/goodside/status/1568375802903015425

299. ​ https://x.com/goodside/status/1568416130133368835

300. ​ https://x.com/goodside/status/1568448128495534081

301. ​ https://x.com/goodside/status/1581868987952300032

302. ​ https://x.com/goodside/status/1612017392518840320

303. ​ https://x.com/goodside/status/1635711013566795776

304. ​ https://x.com/jd_pressman/status/1646766004637401088

305. ​ https://x.com/jeremyphoward/status/1801037736968913128

306. ​ https://x.com/jmilldotdev/status/1592288240861839360

307. ​ https://x.com/lemonodor/status/1628270074074398720

308. ​ https://x.com/littmath/status/1598128056874721283

309. ​ https://x.com/mbusigin/status/1789334007047455178

310. ​ https://x.com/md_rumpf/status/1647911393796956162

311. ​ https://x.com/peterwildeford/status/1522633978305560576

312. ​ https://x.com/shinboson/status/1805459742518595585

313. ​ https://x.com/wgussml/status/1834712489822765295

314. ​ https://x.com/yoheinakajima/status/1670557048743010305

315. ​ https://yaofu.notion.site/A-Closer-Look-at-Large-Language-Models-Emergent-Abilities-493876b55df5479d80686f68a1abd72f

316. ​ 3c24293d28deb791cf7f4769df319264cbc93fb1.html

317. ​ Mind Your Step (by Step): Chain-of-Thought can Reduce Performance on Tasks where Thinking Makes Humans Worse

318. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2410.21333.html

319. ​ Can Long-Context Language Models Subsume Retrieval, RAG, SQL, and More?

320. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2406.13121%2523google.html

321. ​ Intelligent Go-Explore (IGE): Standing on the Shoulders of Giant Foundation Models

322. ​ Jeff Clune—Professor—Computer Science—University of British Columbia

323. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2405.15143.html

324. ​ From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Step

325. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2405.14838.html

326. ​ Retrieval Head Mechanistically Explains Long-Context Factuality

327. ​ Yizhong Wang—University of Washington

328. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2404.15574.html

329. ​ Let’s Think Dot by Dot: Hidden Computation in Transformer Language Models

330. ​ Sam Bowman

331. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2404.15758.html

332. ​ Re-evaluating GPT-4’s bar exam performance

333. ​ https%253A%252F%252Flink.springer.com%252Farticle%252F10.1007%252Fs10506-024-09396-9.html

334. ​ Long-form factuality in large language models

335. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2403.18802%2523deepmind.html

336. ​ Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking

337. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2403.09629.html

338. ​ Tokenization counts: the impact of tokenization on arithmetic in frontier LLMs

339. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2402.14903.html

340. ​ Why are Sensitive Functions Hard for Transformers?

341. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2402.09963.html

342. ​ More Agents Is All You Need

343. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2402.05120%2523tencent.html

344. ​ Beyond Human Data: Scaling Self-Training for Problem-Solving with Language Models (ReST^EM)

345. ​ Abhishek Kumar

346. ​ Igor Mordatch

347. ​ Behnam Neyshabur

348. ​ Jascha Sohl-Dickstein

349. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2312.06585%2523deepmind.html

350. ​ Can Generalist Foundation Models Outcompete Special-Purpose Tuning? Case Study in Medicine

351. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2311.16452%2523microsoft.html

352. ​ Training Chain-of-Thought via Latent-Variable Inference

353. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2312.02179.html

354. ​ Can GPT models be Financial Analysts? An Evaluation of ChatGPT and GPT-4 on mock CFA Exams

355. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2310.08678.html

356. ​ Language Agent Tree Search Unifies Reasoning Acting and Planning in Language Models

357. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2310.04406.html

358. ​ Think before you speak: Training Language Models With Pause Tokens

359. ​ Sanjiv Kumar

360. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2310.02226.html

361. ​ Contrastive Decoding Improves Reasoning in Large Language Models

362. ​ Mike Lewis

363. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2309.09117%2523facebook.html

364. ​ Re-Reading Improves Reasoning in Large Language Models

365. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2309.06275.html

366. ​ From Sparse to Dense: GPT-4 Summarization with Chain of Density (CoD) Prompting

367. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2309.04269.html

368. ​ Solving Challenging Math Word Problems Using GPT-4 Code Interpreter with Code-based Self-Verification

369. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2308.07921.html

370. ​ Unleashing the Emergent Cognitive Synergy in Large Language Models: A Task-Solving Agent through Multi-Persona Self-Collaboration

371. ​ Furu Wei

372. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2307.05300%2523microsoft.html

373. ​ Teaching Arithmetic to Small Transformers

374. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2307.03381.html

375. ​ Let’s Do a Thought Experiment: Using Counterfactuals to Improve Moral Reasoning

376. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2306.14308%2523google.html

377. ​ Thought Cloning: Learning to Think while Acting by Imitating Human Thinking

378. ​ Jeff Clune—Professor—Computer Science—University of British Columbia

379. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2306.00323.html

380. ​ Let’s Verify Step by Step

381. ​ Jan Leike

382. ​ John Schulman’s Homepage

383. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2305.20050%2523openai.html

384. ​ How Language Model Hallucinations Can Snowball

385. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2305.13534.html

386. ​ Tree of Thoughts (ToT): Deliberate Problem Solving with Large Language Models

387. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2305.10601%2523deepmind.html

388. ​ Language Models Don’t Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting

389. ​ Julian Michael

390. ​ Sam Bowman

391. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2305.04388.html

392. ​ Distilling Step-by-Step! Outperforming Larger Language Models with Less Training Data and Smaller Model Sizes

393. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2305.02301%2523google.html

394. ​ Boosting Theory-of-Mind Performance in Large Language Models via Prompting

395. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2304.11490.html

396. ​ How well do Large Language Models perform in Arithmetic tasks?

397. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2304.02015%2523alibaba.html

398. ​ ChatGPT Goes to Law School

399. ​ https%253A%252F%252Fpapers.ssrn.com%252Fsol3%252Fpapers.cfm%253Fabstract_id%253D4335905.html

400. ​ Large Language Models as Fiduciaries: A Case Study Toward Robustly Communicating With Artificial Intelligence Through Legal Standards

401. ​ https%253A%252F%252Fpapers.ssrn.com%252Fsol3%252Fpapers.cfm%253Fabstract_id%253D4335945.html

402. ​ Iterated Decomposition: Improving Science Q&A by Supervising Reasoning Processes

403. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2301.01751%2523elicit.html

404. ​ U-PaLM: Transcending Scaling Laws with 0.1% Extra Compute

405. ​ Yi Tay

406. ​ Jason Wei

407. ​ Neil Houlsby

408. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2210.11399%2523google.html

409. ​ Large Language Models Can Self-Improve

410. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2210.11610%2523google.html

411. ​ Challenging BIG-Bench Tasks (BBH) and Whether Chain-of-Thought Can Solve Them

412. ​ Yi Tay

413. ​ Jason Wei

414. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2210.09261%2523google.html

415. ​ Self-Ask: Measuring and Narrowing the Compositionality Gap in Language Models (Bamboogle)

416. ​ Noah A. Smith

417. ​ Mike Lewis

418. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2210.03350%2523allen.html

419. ​ Language Models are Multilingual Chain-of-Thought Reasoners

420. ​ Yi Tay

421. ​ Jason Wei

422. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2210.03057%2523google.html

423. ​ FOLIO: Natural Language Reasoning with First-Order Logic

424. ​ Caiming Xiong—Home Page

425. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2209.00840.html

426. ​ Can large language models reason about medical questions?

427. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2207.08143.html

428. ​ Inner Monologue: Embodied Reasoning through Planning with Language Models

429. ​ Igor Mordatch

430. ​ Sergey Levine

431. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2207.05608%2523google.html

432. ​ Language Models (Mostly) Know What They Know

433. ​ Saurav Kadavath

434. ​ About Me

435. ​ Andy Jones

436. ​ Sam Bowman

437. ​ https://jack-clark.net/about/

438. ​ Sam McCandlish

439. ​ Jared Kaplan

440. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2207.05221%2523anthropic.html

441. ​ Least-to-Most Prompting Enables Complex Reasoning in Large Language Models

442. ​ Jason Wei

443. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2205.10625%2523google.html

444. ​ Dialog Inpainting: Turning Documents into Dialogues

445. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2205.09073%2523google.html

446. ​ Unifying Language Learning Paradigms

447. ​ Yi Tay

448. ​ Neil Houlsby

449. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2205.05131%2523google.html

450. ​ Socratic Models: Composing Zero-Shot Multimodal Reasoning with Language

451. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2204.00598%2523google.html

452. ​ Self-Consistency Improves Chain-of-Thought Reasoning in Language Models

453. ​ Jason Wei

454. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2203.11171%2523google.html

455. ​ Chain-of-Thought Prompting Elicits Reasoning in Large Language Models

456. ​ Jason Wei

457. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2201.11903%2523google.html

458. ​ Reasoning Like Program Executors

459. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2201.11473%2523microsoft.html

460. ​ A Neural Network Solves and Generates Mathematics Problems by Program Synthesis: Calculus, Differential Equations, Linear Algebra, and More

461. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2112.15594.html

462. ​ WebGPT: Improving the factual accuracy of language models through web browsing

463. ​ Jacob Hilton's Homepage

464. ​ John Schulman’s Homepage

465. ​ https%253A%252F%252Fopenai.com%252Fresearch%252Fwebgpt.html

466. ​ Training Verifiers to Solve Math Word Problems

467. ​ Jacob Hilton's Homepage

468. ​ John Schulman’s Homepage

469. ​ https%253A%252F%252Farxiv.org%252Fabs%252F2110.14168%2523openai.html

470. ​ Decision Transformer: Reinforcement Learning via Sequence Modeling

471. ​ Aravind Rajeswaran

472. ​ Kimin Lee

473. ​ Aditya Grover

474. ​ Michael (misha) Laskin

475. ​ Aravind Srinivas

476. ​ Igor Mordatch

477. ​ https%253A%252F%252Fsites.google.com%252Fberkeley.edu%252Fdecision-transformer.html

478. ​ Word in Context: Agent and Agent Clarification (69% Dev)

479. ​ https%253A%252F%252Fgptprompts.wikidot.com%252Flinguistics%253Aword-in-context%2523toc3.html

480. ​ I found that getting GPT-3 to add its own "internal monologue" in parentheses to be a helpful strategy…

481. ​ https%253A%252F%252Fnews.ycombinator.com%252Fitem%253Fid%253D23990902.html

482. ​ Seems to work

483. ​ https%253A%252F%252Fx.com%252Fkleptid%252Fstatus%252F1284069270603866113.html

484. ​ Teaching GPT-3 to do a brute force 'for loop' checking answers also seems to work

485. ​ https%253A%252F%252Fx.com%252Fkleptid%252Fstatus%252F1284098635689611264.html

486. ​ Inducing Self-Explanation: a Meta-Analysis

487. ​ %252Fdoc%252Fpsychology%252Fspaced-repetition%252F2018-bisra.pdf.html

488. Wikipedia Bibliography:

     1. ​ Language of Thought Hypothesis

     2. ​ Retrospective Think Aloud

     3. ​ Rubber Duck Debugging

     4. ​ Think Aloud Protocol

     5. ​ Thomas L. Griffiths

     6. ​ Anthony Chen

     7. ​ Yejin Choi

     8. ​ Quoc V. Le

     9. ​ Michael Hahn

     10. ​ Qin Zhang

     11. ​ Jeffrey Pennington

     12. ​ Kathleen Kenealy

     13. ​ Tao Qin

     14. ​ Eric Horvitz

     15. ​ Sholto Douglas

     16. ​ Can Xu

     17. ​ Noémie Elhadad

     18. ​ Ke Wang

     19. ​ Heng Ji

     20. ​ Ilya Sutskever

     21. ​ Noah Smith (writer)

     22. ​ Yuan Cao

     23. ​ Jiawei Han

     24. ​ Ed H. Chi

     25. ​ Ludwig Schmidt

     26. ​ Brian Wong

     27. ​ Dragomir Radev

     28. ​ Fei Xia

     29. ​ Scott Johnston

     30. ​ Danny Hernandez

     31. ​ Dario Amodei

     32. ​ Ed Chi

     33. ​ Qian Liu

     34. ​ Nakul Verma

     35. ​ Eugene Wu

     36. ​ Gilbert Strang

     37. ​ Pieter Abbeel