groundy

models & research

125 articles·rss

Top in models & research

models

Analytic Inference Cuts Bayesian Deep Ensemble Serving Cost, But Leaves Training as the Bottleneck

A new arXiv preprint replaces sampling-based averaging in Bayesian deep ensembles with closed-form Bayesian aggregation, cutting per-query inference cost and shifting the.

jul 09
models

Tree-of-Thoughts Improves Text-to-Image Prompting by Reasoning Over Hypotheses, Not Pixels

A new arXiv paper ports Tree-of-Thoughts prompting to text-to-image in-context learning and reports CoBSAT gains, but reasoning runs on prompt hypotheses, not image states.

jul 09
models

FourierQK's spectral Q/K filter cuts TinyShakespeare loss by 79%, but long-context proof is missing

FourierQK filters query and key projections before attention, cutting TinyShakespeare character-level loss by 79%, but word-level, retrieval and long-context tests are absent.

models

Tencent Hunyuan 3's Agent Push Has No Public DeepSeek or Qwen Benchmarks Yet

Tencent's Hy3 is billed as a 295B agent model with a 21B active token footprint, but public materials omit benchmark tables and independent DeepSeek or Qwen comparisons.

models

When Does Memory, Not Compute, Decide Who Can Profitably Serve LLMs?

A July 2026 arXiv paper argues that scarce HBM and DRAM bandwidth, not raw compute, will determine which labs and providers can profitably serve large language models through.

models

Can We Trust LLM Logic? A Graph-Based Stress Test Finds Three Failure Modes

A July 2026 arXiv paper shows Self-Consistency voting can hide contradictory reasoning, and GraphEVAL's graph-based coherence metrics catch flawed paths output checks miss.

models

LongCat-2.0 Hits Claude Opus 4.6 Class on Agents From a 50K-GPU Cluster

LongCat-2.0 is a 1.6T MoE trained on a 50,000-card domestic cluster without NVIDIA silicon, posting agent scores near Claude Opus 4.6 yet CUDA serving stacks still dominate.

models

When Do Time Series Foundation Models Pay Off? The Break-Even Threshold

A July 2026 arXiv break-even study finds pretrained time series foundation models win unconditionally on half of 30 datasets, but lose to ARIMA or XGBoost on a fifth at any.

more in this beat1–50 of 117 · page 1 / 3
  1. jul 06modelsVLA Grounder Tests Language Conditioning to Optimize Black-Box Vision-Action Models
  2. jul 06modelsBlack-Box LLM Architecture Inference: What API Restrictions Reveal About Hidden Model Structure
  3. jul 06modelsInduceKV Tests Continual Learning for Multimodal LLMs Without Expanding Cache
  4. jul 02modelsSonnet 5 vs GPT-5.5: Pricing, Benchmarks, and the Switching Math
  5. jun 29modelsHow LLMs Fuse Conflicting Facts: Single-Source vs Multi-Source Truth
  6. jun 28modelsLinear Transformers Get a Learnable Kernel: Does Flexformer Change the Efficiency Tradeoff?
  7. jun 28modelsHuawei Ships CUDA-Free AI Compute On-Device, but Ascend Quantization Accuracy Is Unverified
  8. jun 28modelsDo Multimodal RAG Models Ignore Late Evidence? A Primacy Bias Test
  9. jun 28modelsCan Deep Learning Design RF Power Amplifiers Without Full EM Simulation?
  10. jun 27modelsSynthetic Clinical Notes from LLMs: Believable Prose Is Not Clinical Validity
  11. jun 27modelsDoubao vs Qwen 3.7 vs GLM-5.2: Route by Axis, Not Leaderboard
  12. jun 27modelsCan Dynamic Experts Fix Catastrophic Forgetting in Robot Manipulation?
  13. jun 27modelsError-Conditioned Neural Solvers vs Iterative Refinement: When Does Learned Correction Win?
  14. jun 27modelsVision-Language Models Move Past Object Detection: The MLLM Perception Shift
  15. jun 27modelsCan Autoregressive Boltzmann Generators Replace MCMC in Simulation?
  16. jun 27modelsLook-Before-Move Plans Observation Before Motion in Dynamic 3D Story Worlds
  17. jun 26modelsGLM 5.2, Qwen 3.7, and DeepSeek in 2026: A Routing Map by Workload, Not by Rank
  18. jun 26modelsSLM Pipeline Catches 10% of Papers Human Reviewers Missed, but No Model Matched Human Accuracy
  19. jun 26modelsMiniMax M3 vs GLM-5.2: Whose 1M-Context Claim Holds Up?
  20. jun 26modelsDeepSeek V4.1 Flash vs Qwen 3.7 vs Llama 4.5: June 2026 HF Trending Ranks Velocity, Not Installs
  21. jun 26models125 Targeted Wikipedia Edits Left a Detectable Signal in Llama Pretraining
  22. jun 26modelsCan SAE Features Stop LLMs From Forgetting During Continual Learning?
  23. jun 26modelsCan a 30B Model Post-Train Itself? A-Evolve-Training Tests Autonomous RL
  24. jun 25modelsOpen-Weight LLM Leaderboards 2026: Where DeepSeek, Qwen, and GLM Rank
  25. jun 25modelsQwen3.7-Max's Top-Ranked Claim vs the Artificial Analysis Index
  26. jun 25modelsDoes Tree-of-Thought Reasoning Scale to Billion-User Modeling?
  27. jun 25modelsCan LLMs Debug Verilog? VeriPilot Puts an Agent on RTL Errors
  28. jun 24modelsTask Decomposition Helps LLMs by Shrinking Output Space, Not by Cutting Labeling Cost
  29. jun 24modelsFlow Matching vs U-Net: A Skip-Free Backbone for Speech Models
  30. jun 24modelsA Per-Neuron Sequence Model Was Withdrawn From arXiv as Coverage Hailed It
  31. jun 24modelsPV-TAM Corrects Decoding Drift and Boundary-Marker Bias in VLM Localization Scoring
  32. jun 24modelsMeituan's General 365 Benchmark: Top Models All Score Under 63%
  33. jun 24modelsLLM Surrogates in A/B Tests: The 39% Recovery Gap and the Silent Bias Risk
  34. jun 24modelsLLM Token Pricing vs Compute Cost: What the Tokenomics Math Shows
  35. jun 24modelsDo LLM Judges Favor Their Own Output? A Sanity Check on Self-Preference
  36. jun 23modelsCan AI Write CAD Programs? CADBench Measures the Gap
  37. jun 23modelsByteDance's Doubao 2.1 Pro vs GPT-5.5: Reading Self-Reported Benchmarks
  38. jun 19modelsCan RoboSSM's State-Space Backbone Replace Transformer Imitation Policies?
  39. jun 19modelsPruning Experts to Shrink MoE Models: Does Attribution-Guided Compression Beat Magnitude?
  40. jun 19modelsGLM-5.2 vs Kimi K2.7 Code: Two Open-Weight Bets on Agentic Coding
  41. jun 19modelsHow Linear Is a Transformer Feed-Forward Block? A New Test Says It's Learned, Not Built In
  42. jun 18modelsGLM-5.2 Benchmarks: What 62.1% SWE-bench Pro and 99.2% AIME Actually Mean
  43. jun 18modelsGLM-5.2 on Terminal-Bench 2.1: Strengths, Gaps, and How to Route Real Coding Tasks
  44. jun 18modelsGLM-5.2 vs Claude Opus 4.8: Open-Weight Coding at Frontier Pricing
  45. jun 18modelsGLM-5.2's 753B MoE Costs More to Self-Host Than the MIT License Suggests
  46. jun 17modelsSTAR Replaces Scalar Reward in Text-to-Image RL with Attention-Derived Spatial Maps
  47. jun 15modelsCan Editing One Neuron Fix LLM Repetition Loops?
  48. jun 10modelsClaude Fable 5 Benchmarks: What FrontierCode, CursorBench, and ViBench Show
  49. jun 10modelsDoes Attribution Patching Lie? A Fix for a Common Interpretability Shortcut
  50. jun 11modelsCan You Make a Multimodal Model Unlearn With Activation Steering?
about this beateditorial framing

Every new foundation model arrives wrapped in a benchmark chart and a press release. This beat exists because the chart almost never tells you what the model actually does — and the press release tells you even less. The interesting work is one layer down: the attention variant that changes the cost curve, the parameterization fix that makes scaling laws actually transfer, the eval protocol whose failure modes flip the standings once you change a decoding parameter or a codec. We cover that layer.

We treat open-weight and closed releases as the same story told from opposite ends of a distribution shift. A trillion-parameter mixture beating a dense competitor on one harness and losing on another is not a contradiction; it is evidence about what the harness measures. Training-efficiency claims, context-window claims, and reasoning-benchmark claims all get the same treatment — read the method section, find the assumption that was load-bearing, and report whether removing it changes the result. When a paper proposes a new safety mechanism, a new compression trick, or a new continual-learning split, we are interested in the part the authors did not want to highlight.

The throughline is comparative and skeptical without being contrarian. Foundation-model research has become the field where the largest gap between published numbers and deployed behavior tends to live. Closing that gap — with reproduction notes, ablation reading, and honest accounting of what generalizes versus what was tuned into the eval — is the beat.