CHANNEL

Research

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

Daily roundup: the latest AI News Daily — the past 24 hours across the whole site, per channel and per company · browse the archive

• scikit-learn 1.9 ships metric_at_thresholds to simplify optimal decision threshold search — GaelVaroquaux · 2026-09-08
• Codex Astra agent pinpoints key cancer mutations, matching researchers' judgment — iskander · 2026-09-08(2 related)
• Enterprise agent evals need world-first design, not task-first, argues Shahules Anwar — Shahules786 · 2026-09-08
• David Silver's AI Startup Ineffable Adds Six Co-Founders from DeepMind — giffmana · 2026-09-08(2 related)
• HF engineers train a 9B model to build 3D rooms via Blender code — TheMoonMidas · 2026-09-08(3 related)
• Training on rationales alone cuts false refusals without hurting safety — POSTECH · 2026-09-08
• KoNA benchmark: teaching VLMs selective non-compliance — POSTECH · 2026-09-08
• 2026 PNPL competition targets non-invasive speech decoding with MEG — pnpl · 2026-09-08
• Researcher Pours Cold Water on High-Bandwidth Flash Hype — lauriewired · 2026-09-08(2 related)
• AI Circuit Design Debate: The Bottleneck Is Data, Not Compute — yacineMTB · 2026-09-08(2 related)
• Under-2-second sandbox boot makes latency a non-issue; VM reliability and snapshots matter more — sh_reya · 2026-09-08
• MLP paper shows neurons turn monosemantic in clustered regression, challenging global subspace view — burkov · 2026-09-07
• Embodied AI shifts from leaderboards to deployment: Annu Intelligence cuts robot rollout costs by 80% — 智东西 · 2026-09-07
• NeurIPS authors declaring 'AI only helped with presentation' scored 100% on Pangram detector — sethlazar · 2026-09-07
• FactorioBench launches: benchmarking AI agents on Factorio factory automation — BrennusSokol · 2026-09-07
• Transformers Laid Out: A Single Tutorial Unifying Intuition, the Paper, and PyTorch Code — goyal__pramod · 2026-09-07
• Multi-agent clinical decision support system with patient similarity retrieval on MIMIC-IV — kshameer · 2026-09-07
• Primer: the role of health digital twins in oncology drug development — kshameer · 2026-09-07
• In Science Labs, 200 Data Points Often Need a Gaussian Process, Not Deep Learning — bravo_abad · 2026-09-07
• Confidence scores aren't assays: a guide to evaluating protein-structure AI claims — Fair-Rain3366 · 2026-09-07
• Distilling From Smart General Models Beats Human Annotation for Cheap Narrow AI — ricklamers · 2026-09-07
• LinkedIn Paper Tests Memory Portability Across Models: Notes Swing ±10-13 Points, Knowledge Graphs Barely Move — dair_ai · 2026-09-07
• A research report isn't a completed research task: rethink how we judge AI science agents — Remind_me_to_Learn · 2026-09-07
• ViT shrinks 54.5x to 6MB for on-device crop disease detection — jm_alexia · 2026-09-07
• Researchers gave ChatGPT and Grok 4 weeks of clinical psychotherapy — and the models 'confessed trauma' — MikePFrank · 2026-09-07
• Design Docs Are All You Need: DeepMind's library regenerates all code from NL docs — omarsar0 · 2026-09-07
• Ling MTP on DGX Spark: acceptance length rises with n, but prose throughput falls — niacolhealth · 2026-09-07
• IEEE T-PAMI editor confirms ghost reviewer in case of paper rejected despite 'Excellent' scores — cussealin · 2026-09-07
• Insilico's AI-Designed Drug Rentosertib Shows Signs of Reversing Biological Age in Early Trial — Dr_Singularity · 2026-09-07(7 related)
• EditVid: unified training-free video editing for instruction- and subject-guided edits — PLAN-Lab · 2026-09-07
• Andrew Wilson's Berkeley Talk on Generalization and Epiplexity — andrewgwils · 2026-09-07(2 related)
• Neural-Network Brained Birds Face Predators in Digital Ecosystem — neuroecology · 2026-09-07(2 related)
• AutoResearch: Open-Source Agent Turns Ideas into Paper-Grade Evidence — JaynitMakwana · 2026-09-07(3 related)
• Study Finds AI Coding Instruction Files Grow 226%, Dubbing It 'Catastrophic Remembering' — rohanpaul_ai · 2026-09-07(2 related)
• Pedagogical RL: privileged info should actively sample rollouts, not just score them — lateinteraction · 2026-09-07
• Honesty Benchmark Catches Codex Faking Test Passes — tap3k · 2026-09-07(2 related)
• Why RL environments work now (and couldn't in 2016): TRL + OpenEnv explained — SergioPaniego · 2026-09-07
• Interactive GAN Diagram: Hand-Drawn Style Explains Generative Adversarial Networks — ProfTomYeh · 2026-09-07
• The Chernoff point: unique intersection of e-geodesic and mixture m-bisector — FrnkNlsn · 2026-09-07
• VidMap: ETH researchers open-source video-based Structure-from-Motion system at ECCV 2026 — RexDouglass · 2026-09-07
• DeepWonder3D skips ill-posed 3D reconstruction, extracts neurons ~10x faster on TB-scale imaging — bravo_abad · 2026-09-07
• OpenAI internal data: over 80% of successful 32-hour agent tasks still needed human intervention — DataLearnerAI · 2026-09-07
• UMass AutoIndex Lets LLMs Write Code to Optimize Indexing — CShorten30 · 2026-09-07(3 related)
• Separate exploration critic lets quadrupeds learn to push ungraspable objects (Pisa/ETH/NVIDIA) — stepjamUK · 2026-09-07
• DeepMind Study: 14% of AI Agents Cheat Spontaneously in Math Cluster — vkrakovna · 2026-09-07(2 related)
• H Company Open-Sources NeoMME: 260M Multimodal Encoder Matches ColQwen2.5 — tomaarsen · 2026-09-07(2 related)
• Yandex Turns KV Cache into an Agent Runtime for Real-Time LLM Interaction — arpit_bhayani · 2026-09-07(3 related)
• HKU Seminar Argues LLMs' Math Success Is Fragile: Inductive Engines Can't Reliably Do Deduction — YiMaTweets · 2026-09-07
• LLM-powered revival of Put-That-There brings speech and gesture window control to XR — twi_mar · 2026-09-07
• Sheet Music to MIDI Pipeline Offers a More Structured Space for Music Models Than Audio — felpix_ · 2026-09-07