CHANNEL

Research

"Research" is a topic channel on AGI Hunt, an AI news site updated around the clock every 30 minutes. Coverage: Papers, new methods, empirical findings, technical and research reports, academia (incl. AI4Science); benchmark and dataset releases land here.

• New AI economy analysis is based on 15 million de-identified human-AI interactions — JMateosGarcia · 2026-07-23
• AI voice phishing can match human scammers while costing far less — MyFest · 2026-07-23
• AI Scientists Conference 2026 Adopts AI-Human Review Process — paraschopra · 2026-07-23(2 related)
• COLMAP merges DEGENSAC into main RANSAC, boosting plane-heavy scenes — ducha_aiki · 2026-07-23
• ByteDance's WorkflowGym Exposes GUI Agents: Top Models Pass Only 30% — 机器之心 · 2026-07-23
• Physics Knowledge in LLMs Lies in Hidden State Transformations — ProfBuehlerMIT · 2026-07-23(2 related)
• Inkling AI Explainer Focuses on Architecture, Training and Benchmarks — MoonL88537 · 2026-07-23
• Pi exposes cache behavior as a debate over agent harnesses burning KV caches heats up — mitsuhiko · 2026-07-23
• TPAMI paper unifies Bregman divergences with surrogate losses for learning — FrnkNlsn · 2026-07-23
• Nathan Benaich says AI will design nearly every molecule soon — nathanbenaich · 2026-07-23
• New Benchmark Shows Top VLMs Struggle with ASCII Diagrams — East-Muffin-6472 · 2026-07-23(3 related)
• DocOps benchmark finds frontier agents still fail on long-horizon document tasks — Jiazhen Jiang · 2026-07-23
• Stanford’s vine-like soft robot grows from the tip to reach trapped people — lukas_m_ziegler · 2026-07-23
• First CAR-T Cell Therapy Approved for Solid Tumors in Gastric Cancer — Dr_Singularity · 2026-07-23
• A production multi-agent team says deterministic orchestration works better than deterministic LLMs — njanChe1 · 2026-07-23
• LxMLS 2026 shares a public video-lecture collection from Lisbon Machine Learning School — caglar_ee · 2026-07-23
• WSD Schedule vs. Early Stopping: Open Speech Model Training Post-mortem — irombie · 2026-07-23
• A video says an economic model points to possible AI-driven knowledge collapse — Tasty-Aspect-6936 · 2026-07-23
• TinyAya v0.3 open speech-to-speech model adds text supervision and 36-layer LoRA — irombie · 2026-07-23
• DeepMind says video generators can become general-purpose vision models — jiqizhixin · 2026-07-23
• Open-source Aster proposes cost-staged verification for LLM code review — letandrewcook · 2026-07-23
• Paper evaluates speech translation by whether listeners can answer key questions — EhudReiter · 2026-07-23
• Alignment paper warns static preferences can lock in values and stall societies — edelwax · 2026-07-23
• LLMs may help most in the manual refinement phase of decompilation — OwariDa · 2026-07-23
• RLSS 2026 Milan MasterClass Slides & Video Released — misovalko · 2026-07-23(5 related)
• Ascend SuperPOD Optimization Boosts DeepSeek-V4 Training MFU to 34.22% — pmttyji · 2026-07-23(2 related)
• Rethinking Diffusion Models: Outdated Metrics and Forced Architecture Coupling — kalomaze · 2026-07-23(7 related)
• HPD-Parsing Achieves High-Throughput Document Parsing via Hierarchical Parallel Decoding — pmttyji · 2026-07-23(2 related)
• AlbumentationsX MCP lets AI assistants preview and edit augmentation pipelines locally — viglovikov · 2026-07-23
• MetaCircle proposes a non-LLM “Mechanistic Auto-Research” system — ZimingLiu11 · 2026-07-23
• HiSME lets LLM agents evolve the way they evolve skills, without changing weights — 机器之心 · 2026-07-23
• Hypernetworks show power-law scaling for factual knowledge injection in LLMs — Nischay Dhankhar · 2026-07-23
• Open-Sourced DSpark Speculator Boosts Inkling Throughput by 1.89x — ying11231 · 2026-07-23(3 related)
• New atlas maps 2,226 coding tasks across 11 benchmarks to expose coverage gaps — zainhas · 2026-07-23
• America’s first Distillation Summit will cover RL, agents, IP and national security — AkshatS07 · 2026-07-23
• Symbolic algebra check suggests a possible Jacobian conjecture counterexample in C^3 — sloppenheimer · 2026-07-23
• Surge in AI Math Proofs Signals Imminent Breakthroughs in Other Fields — emollick · 2026-07-23
• Optical phase-change memory shows 40 dB loss but a 2D VCSEL array — jwt0625 · 2026-07-23
• RECAP lets probes verify activation explanations that reconstruction scores can fake — Hiskias Dingeto · 2026-07-23
• USC benchmark shows GPT-5.5 scores 10.6% on active visual observation tasks — UniversityofSouthernCalifornia · 2026-07-23
• Paper proposes Linearizing Softmax Attention into Gated DeltaNet — bronzeagepapi · 2026-07-23
• Overcoming RGB Sim-to-Real Gap: Gaussian Splatting Enables Zero-Shot Robot Transfer — ZeYanjie · 2026-07-23
• Study Reveals Long-Context Models Waste Tokens Copying Text — rohanpaul_ai · 2026-07-23(2 related)
• A math meme says you’re behind if you aren’t using LLMs to disprove conjectures — MickeySteamboat · 2026-07-23
• A curated guide to LLM cache management spans KV cache, batching, and decoding — gaganghotra_ · 2026-07-23
• ICLR 2024 paper uses LLMs to accelerate RL on 25+ long-horizon robot tasks — rsalakhu · 2026-07-23
• Kimi K3 finds 32 of 86 vulnerabilities for just $19.37 in benchmark run — zeeg · 2026-07-23
• AI Content Detectors No Better Than a Coin Toss — burkov · 2026-07-23(2 related)
• CoLT teaches multimodal models to think in latent steps and cuts inference 10.1x — jiqizhixin · 2026-07-23
• Duke study says better memory, not a bigger model, lifted ARC-AGI-3 scores by 18 points — imjustnewatai · 2026-07-23