groundy
modelspage 2 / 3· back to top →

models & research

more in this beat51–100 of 117 · page 2 / 3
  1. jun 11modelsWhy Pruning a Model Can Raise Its Out-of-Distribution Accuracy
  2. jun 09modelsDo Unified Multimodal Models Actually Interleave Understanding and Generation?
  3. jun 09modelsHow LLMs Track Who Did What: The Entity Rebinding Circuit
  4. jun 09modelsClaude Fable 5 vs Opus 4.8: When 2x Pricing Is Worth It
  5. jun 09modelsClaude Mythos 5 Access Rules: Who Gets Project Glasswing and Why
  6. jun 09modelsFable 5 Distillation Protection: How Anthropic Blocks Model Copying
  7. jun 09modelsSkip Fable 5 or Upgrade? When Opus 4.8 and Sonnet 4.6 Are Still Enough
  8. jun 08modelsLLM Steganography: Can Defenders Detect Payloads Hidden in Model Output?
  9. jun 08modelsDo Privacy Defenses Actually Protect Fine-Tuned LLMs? A New Benchmark
  10. jun 08modelsCan You Reconstruct an LLM's System Prompt From Its Activations?
  11. jun 08modelsDoes Softmax Normalization Limit What Attention Can Represent?
  12. jun 07modelsCan an Attacker Steal Your Model's Last Layer From Its Outputs?
  13. jun 06modelsCan LLMs Leak Training Data? A New Test Splits Capacity From Intent
  14. jun 06modelsWhen an AI Agent's Tools Break, Can It Recover? A New Benchmark
  15. jun 05modelsMiniMax M3 Bets on Sparse Attention for 1M Context. Does the Math Hold?
  16. jun 05modelsCan One Model Handle Every CAD Task? UniCAD Tests It
  17. jun 05modelsDo Foundation Models Actually Learn Relational Structure In-Context?
  18. jun 05modelsCan LLMs Write Better Research Paper Titles Than Authors?
  19. jun 05modelsDoes Information-Theoretic Example Selection Beat kNN for In-Context Learning?
  20. jun 05modelsDo Concept Bottleneck Model Benchmarks Measure Interpretability or Dataset Bias?
  21. jun 05modelsContinuous Bit-Width Quantization vs Fixed INT4: Does LiftQuant Beat Discrete?
  22. jun 04modelsFederated Learning for Industrial IoT Anomaly Detection: The Data-Locality Tradeoff
  23. jun 04modelsReading Failed LLM Reasoning Traces Won't Tell You Which Ones RL Can Fix
  24. jun 04modelsCan You Stitch Two Foundation Models Together Without Retraining?
  25. jun 04modelsDo Reasoning LLMs Waste Tokens? OckBench Tries to Measure It
  26. jun 03modelsWhich Layer Detects LLM Hallucinations Best? The Case Against Fixed-Layer Probes
  27. jun 02modelsCross-Domain RL Training Degrades Capabilities. CARE-RL Reweights to Fix It
  28. jun 02modelsLLM Watermarking Without Quality Loss: The Non-Distortionary Approach
  29. jun 01modelsTreating LLM Agent Memory as a Database: The VikingMem Approach
  30. jun 01modelsCan a Language Model Work Without a Neural Network? A New arXiv Paper Says Yes
  31. jun 01modelsCan Code-Generating LLMs Do Engineering Math? FEM-Bench Tests Them
  32. jun 01modelsUnlearning Isn't Deletion: arXiv 2505.16831 Shows Machine Unlearning in LLMs Is Reversible
  33. may 31modelsWhy LLMs Fail at Spatial Reasoning When Planning Navigation
  34. may 31modelsDoes Giving AI Agents More Skills Help? A Controlled SkillsBench Study
  35. may 30modelsCan an LLM Peer-Review Your Paper? A New Behavior Benchmark
  36. may 30modelsAnthropic Scaled Sparse Autoencoders to Claude 3 Sonnet. Interpretability Now Costs Compute
  37. may 28modelsTracing Why LLM Agent Memory Fails: A Method for Attributing Errors
  38. may 28modelsPersona Prompts Change Who an LLM Recommends as an Expert
  39. may 27modelsOpus 4.8 vs Opus 4.7: What Changed and What Did Not
  40. may 27modelsOpus 4.8 Batch API: 1M Context, 300k Output, and Team Cost Controls
  41. may 26modelsScale Vectors: Tiny Parameter Subsets That Disproportionately Steer LLM Behavior
  42. may 26modelsOne Learning Rate Doesn't Fit All: Heavy-Tail Layerwise LR Schedules for LLM Pretraining
  43. may 25modelsAudio LLMs Break When the Codec Changes: A Robustness Vector Voice-AI Teams Haven't Tested
  44. may 25modelsDo LLMs Know What Not to Say? Causal Evidence for Statistical Preemption
  45. may 24modelsEmbedding Compression at Training Time: DIVE's Gradient Trick vs Post-Hoc Quantization for Vector DBs
  46. may 24modelsμP Hyperparameter Transfer Has an Embedding Layer Hole, New arXiv Paper Says
  47. may 23modelsProject Glasswing One Month In: AI Bug Discovery Has Outpaced the Patch Pipeline
  48. may 22modelsarXiv 2605.16428 Measures AI Search's Drag on Publisher Traffic Using Paired Google and Reddit Data
  49. may 22modelsA Theory of Time-Sensitive Language Generation Says Sparse Hallucination Beats Mode Collapse
  50. may 18modelsThe Last Word Often Wins: A Format Confound Inflates Chain-of-Thought Corruption Robustness Scores