DataSci Ocean
Posts Tags Categories
English
DataSci Ocean

All Posts

2026

Harness Engineering: Building Self-Evolving AI Agents with Ralph Loop and MemRL 04-05
WebResearcher: Redefining Deep Research through the Art of Iterative State Reconstruction 04-02
Takeaways of vLLM Semantic Router 03-29
Beyond Fine-Tuning: How MemRL Enables Self-Evolving AI Agents via Reinforcement Learning 03-29
UniversalRAG: Mastering Multimodal RAG with Intelligent Routing and Granularity Control 03-25
Beyond the Context Window: A Deep Dive into Context Engineering for AI Agents 03-22
Beyond Prompt Engineering: The System Engineering of AI Agents — Deconstructing Memory, Tool-Making, and Autonomy in OpenClaw 03-16
Why Your RAG Fails at Complex Tables—And How MixRAG Fixes It: A Deep Dive into Heterogeneous Document Retrieval 03-15
Deep Dive into MRAG: Solving Temporal Reasoning in RAG with Symbolic Logic (EMNLP 2025) 03-10
Selective Clarification: Overcoming LLM Hallucinations via the CLAM Framework 02-25
[EMNLP 2025] TableRAG Deep Dive: Text is for Reading, Tables are for Querying (via SQL) 02-13
Stop LLMs From Blindly Guessing: A Deep Dive into "Clarify When Necessary" (NAACL 2025) 02-08
Beyond HyDE: How ReDE-RF Makes RAG 10x Faster by "Judging" Instead of "Writing" 02-03
VideoDR: Bridging the Gap Between Video Understanding and Agentic Search on the Open Web 01-25
Stop Using Giant LLMs for Everything: Why NVIDIA Research Says Small Language Models (SLMs) Are the Future of AI Agents 01-19
Google's "Free Lunch" for LLMs: How Prompt Repetition Fixes Attention Bottlenecks with Zero Latency 01-19
Is RLHF Killing Creativity? How Verbalized Sampling Mitigates Mode Collapse and Unlocks LLM Diversity 01-03
Beyond Self-Consistency: How CER Boosts LLM Reasoning by Leveraging "Process Confidence" (ACL 2025) 01-03

2025

Stop AI Hallucinations Early: How Meta's DeepConf Uses Token Confidence to Cut Inference Costs by 80% 12-28
rStar Deep Dive: How MCTS & Mutual Reasoning Boost LLaMA2-7B Accuracy from 12% to 64% Without Fine-tuning 12-10