Advanced Reasoning Techniques for AGI

In using LlMs, we regularly rely on prompts utilizing Direct Reasoning (DR) for their clear and direct approach to solving problems (example of DR: “If the sky is dark and cloudy, it is likely to rain. Today, the sky is dark and cloudy. Therefore, it is likely to rain today.” ) However, DR has inherent limitations, such as its linear approach, which might not adequately address complex issues that require nuanced understanding and multi-step logical deductions. These shortcomings can lead to an oversimplified analysis, overlooking crucial details and indirect relationships, which might result in less accurate or comprehensive outcomes. This paper introduces an Indirect Reasoning (IR) approach, leveraging contrapositives and contradictions to enhance reasoning in tasks like factual analysis and mathematical proofs. By augmenting rules and employing prompt templates that initiate IR via contradiction proofs, the technique significantly improves LLMs like GPT-3.5-turbo and Gemini-pro in reasoning accuracy. Notably, IR has been shown to boost LLMs' performance by 27.33% in factual reasoning and 31.43% in mathematical proof tasks, marking a substantial advancement in their analytical capabilities. The enhancements in reasoning accuracy achieved through IR suggest that delving into logical equivalences and contradictions offers a promising path for further improving LLMs. This approach not only boosts accuracy in specific tasks but also broadens the models’ applicability across various domains, thereby increasing their overall utility and effectiveness in real-world applications. Here is an example of IR: “Let’s assume that despite the sky being dark and cloudy, it will not rain today. Given this assumption, we would expect no need for umbrellas or raincoats, and outdoor events would proceed without weather-related disruptions. However, if historical patterns and weather forecasts indicate that dark, cloudy skies almost always precede rainfall, this assumption leads to a contradiction with expected outcomes and preparedness for rain. Therefore, the initial assumption must be false, suggesting it is likely to rain today with the dark and cloudy sky.” Paper: https://lnkd.in/eEzsrnrD

Want a prompting technique that is better than RAG or Chain of Thought? Merge them. Chain of Thought is easy. Just add “let’s think step by step” to the instructions. It miraculously encourages the language model to break down its task into subtasks, and gets better results than just asking for the answer. Unfortunately, it also increases the model’s propensity to hallucinate, especially when the “chain” has a lot of links. Retrieval Augmented Generation is the technique of looking up some references and sticking them into the prompt, so that the model can use outside knowledge. You get fewer hallucinations, but also less planning and reasoning than with CoT. A new paper proposes a fusion of these two techniques to get both long term planning and factuality. It apparently works really well, particularly for tasks that require planning and multiple steps: 13.63% improvement on code generation, 16.96% on mathematical reasoning, 19.2% on creative writing, and 42.78% on embodied task planning. Here’s how it works: You start with a Chain of Thought, then you go to Thought #1 and try to validate it with some outside knowledge for it. If it is invalid, you adjust Thought #1. If it is valid, you go on to thought #2, and so on until you are done. The researchers unfortunately named this technique Retrieval Augmented Thoughts (RAT). But we won’t hold that against them… Paper: https://buff.ly/43xuCOD Code: https://buff.ly/4amIrll #ArtificialIntelligence #AIResearch #DeepLearning #NLP #CodeGeneration #MathematicalReasoning #CreativeWriting #TaskPlanning #RetrievalAugmentedGeneration

Prompt engineering is a rapidly-evolving topic in AI, but recent research can be grouped into four categories... (1) Reasoning: Simple prompting techniques are effective for many problems, but more sophisticated strategies are required to solve multi-step reasoning problems. - [1] uses zero-shot CoT prompting to automatically generate problem-solving rationales to use for standard CoT prompting. - [2] selects CoT exemplars based on their complexity (exemplars that have the maximum number of reasoning steps are selected first). - [3] improves CoT prompting by asking the LLM to progressively refine the generated rationale. - [4] decomposes complex tasks into several sub-tasks that can be solved via independent prompts and later aggregated into a final answer. (2) Tool Usage: LLMs are powerful, but they have notable limitations. We can solve many of these limitations by teaching the LLM how to leverage external, specialized tools. - [5, 6] finetune a language model to teach it how to leverage a fixed, simple set of text-based APIs when answering questions. - [7] uses a central LLM-based controller to generate a program—written in natural language—that composes several tools to solve a complex reasoning task. - [8] uses a retrieval-based finetuning technique to teach an LLM to adaptively make calls to APIs based on their documentation when solving a problem. - [9] uses an LLM as a central controller for leveraging a variety of tools in the form of deep learning model APIs. - [10, 11] integrates code-capable LLMs with a sandboxed Python environment to execute programs when solving problems. (3) Context Window: Given the emphasis of recent LLMs on long contexts for RAG / few-shot learning, the properties of context windows and in-context learning have been studied in depth. - [12] shows that including irrelevant context in the LLM’s prompt can drastically deteriorate performance. - [13] finds that LLMs pay the most attention to information at the beginning/end of the prompt, while information placed in the middle of a long context is forgotten. - [14] proposes a theoretically-grounded strategy for optimally selecting few-shot exemplars. (4) Better Writing: One of the most popular use-cases of LLMs is for improving human writing, and prompt engineering can be used to make more effective writing tools with LLMs. - [15] improves the writing abilities of an LLM by first generating an outline and then filling in each component of the outline one-by-one. - [16] uses a smaller LLM to generate a “directional stimulus” (i.e., a textual hint) that can be used as extra context to improve an LLM’s writing ability on a given task. - [17] improves the quality of LLM-generated summaries by iteratively prompting the LLM to increase the information density of the summary.

📝 Announcing our paper that proposes a unified cognitive and computational framework for Artificial General Intelligence (AGI) -- going beyond token-level predictions -- one that emphasizes modular reasoning, memory, agentic behavior, and ethical alignment 🔹 𝐓𝐡𝐢𝐧𝐤𝐢𝐧𝐠 𝐁𝐞𝐲𝐨𝐧𝐝 𝐓𝐨𝐤𝐞𝐧𝐬: 𝐅𝐫𝐨𝐦 𝐁𝐫𝐚𝐢𝐧‑𝐈𝐧𝐬𝐩𝐢𝐫𝐞𝐝 𝐈𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐜𝐞 𝐭𝐨 𝐂𝐨𝐠𝐧𝐢𝐭𝐢𝐯𝐞 𝐅𝐨𝐮𝐧𝐝𝐚𝐭𝐢𝐨𝐧𝐬 𝐟𝐨𝐫 𝐀𝐫𝐭𝐢𝐟𝐢𝐜𝐢𝐚𝐥 𝐆𝐞𝐧𝐞𝐫𝐚𝐥 𝐈𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐜𝐞 𝐚𝐧𝐝 𝐢𝐭𝐬 𝐒𝐨𝐜𝐢𝐞𝐭𝐚𝐥 𝐈𝐦𝐩𝐚𝐜𝐭 🔹 In collaboration with University of Central Florida, Cornell University, UT MD Anderson Cancer Center UTHealth Houston Graduate School of Biomedical Sciences, Toronto Metropolitan University, University of Oxford, Torrens University Australia, Obuda University, Amazon others. 🔹 Paper: https://lnkd.in/gqKUV4Mr ✍🏼 Authors: Rizwan Qureshi, Ranjan Sapkota, Abbas Shah, Amgad Muneer, Anas Zafar, Ashmal Vayani, Maged Shoman, PhD, Abdelrahman Eldaly, Kai Zhang, Ferhat Sadak, Shaina Raza, PhD, Xinqi Fan, Ravid Shwartz Ziv, Hong Yang, Vinija Jain, Aman Chadha, Manoj Karkee, @Jia Wu, Philip Torr, FREng, FRS, Seyedali Mirjalili ➡️ 𝐊𝐞𝐲 𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬 𝐨𝐟 𝐓𝐡𝐢𝐧𝐤𝐢𝐧𝐠 𝐁𝐞𝐲𝐨𝐧𝐝 𝐓𝐨𝐤𝐞𝐧𝐬' 𝐂𝐨𝐠𝐧𝐢𝐭𝐢𝐯𝐞‑𝐂𝐨𝐦𝐩𝐮𝐭𝐚𝐭𝐢𝐨𝐧𝐚𝐥 𝐀𝐆𝐈 𝐅𝐫𝐚𝐦𝐞𝐰𝐨𝐫𝐤: 🧠 𝐅𝐨𝐮𝐧𝐝𝐚𝐭𝐢𝐨𝐧𝐚𝐥 𝐅𝐫𝐚𝐦𝐞𝐰𝐨𝐫𝐤: Integrates cognitive neuroscience, psychology, and AI to define AGI via modular reasoning, persistent memory, agentic behavior, vision-language grounding, and embodied interaction. 🔗 𝐁𝐞𝐲𝐨𝐧𝐝 𝐓𝐨𝐤𝐞𝐧‑𝐏𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐨𝐧: Critiques token-level models like GPT-4.5 and Claude 3.5, advocating for test-time adaptation, dynamic planning, and training-free grounding through retrieval-augmented agentic systems. 🚀 𝐑𝐨𝐚𝐝𝐦𝐚𝐩 𝐚𝐧𝐝 𝐂𝐨𝐧𝐭𝐫𝐢𝐛𝐮𝐭𝐢𝐨𝐧𝐬: Proposes a roadmap for AGI through neuro-symbolic learning, value alignment, multimodal cognition, and cognitive scaffolding for transparent, socially integrated systems.