A robot may know how to move yet still need evidence about what to do. This 100+ page survey organizes the research on supplying that context. Title: In-Context Learning for Robots: Methods and Applications URL: ❓ What is in-context learning for robots? It's adaptation that uses demonstrations and interaction to direct existing competence, with neural parameters held fixed at deployment. Broad pretraining gives a robot reusable motor skill, but the current scene can still leave several procedures feasible, hide an earlier event, or say nothing about how an unfamiliar material responds. More motor competence alone never closes that gap. ❓ How is it different from fine-tuning? Fine-tuning folds new evidence into the weights; in-context learning makes it available at the decision instead. The survey separates five places adaptation can live: supplied context, recurrent state, external memory, test-time weight updates, and persistent weight updates. The first three are the parameter-fixed regime. ❓ How are the methods organized? Into four families, split by the intermediate that execution actually consumes. ・Context-conditioned policies (infer actions straight from the evidence) ・Geometric demonstration transfer (align a reference motion, then track it) ・World-model-based control (predict futures, then plan) ・Skill- and agent-based execution (pick skills or programs for a separate executor) The key point is that one demonstration can feed any of the four; what separates them is the intermediate execution needs, not how the evidence was gathered. ❓ How should this be evaluated? By intervening under matched execution conditions: remove the supplied evidence, clear the memory, suppress retrieval, restore pre-adaptation weights. One reported comparison shows how hard the open end is: asked to discover an unfamiliar action convention unaided, the system succeeded in 1 of 20 episodes and got only 23.3% of the mappings right. #Robotics# #InContextLearning#

🧠 From "reading context" to "skillfully learning from it." This method has an LLM acquire context-specific skills through self-play alone, with no human annotation and no external feedback. Title: From Context to Skills: Can Language Models Learn from Context Skillfully? URL: 📝 Overview LLMs are strong on knowledge seen in pretraining but weak on novel, specialized contexts. This paper proposes Ctx2Skill, which autonomously discovers and refines context-specific skills without human annotation or external feedback. ❓ Challenges Solved Annotating long, technically dense documents is prohibitively costly. And unlike coding, context learning has no execution feedback to verify, making automated skill construction hard. 💡 Methodology & Proposed Approach ・A multi-agent self-play of five frozen-LM roles iterates N=5 times over M=5 tasks ・A Challenger creates tasks and rubrics probing weaknesses, a Reasoner solves them, and a Judge gives pass/fail verdicts ・Proposer and Generator pairs diagnose failures and synthesize skill updates ・A Cross-Time Replay mechanism maximizes the product of hard- and easy-probe performance to pick the most generalizable skill set across iterations 🎯 Use Cases It fits feeding a model long specialized documents and having it acquire the needed skills on the spot, directly useful where you need rapid adaptation to domain-specific knowledge. 📊 Experimental Results ・On CL-Bench (500 contexts, 1,899 tasks), GPT-4.1's solving rate rose from 11.1% to 16.5% ・GPT-5.1 went from 21.1% to 25.8%, and GPT-5.2 from 18.2% to 21.4% ・Skills transfer from stronger to weaker models: GPT-5.1's skills on GPT-4.1 give 16.1% ・The augmented GPT-4.1 (16.5%) beats an unaugmented Gemini 3 Pro (15.8%) #LLM# #InContextLearning#