Why Far Looks Up: Probing Spatial Representation in Vision-Language Models
Paper • 2605.30161 • Published • 38 •
3
Noah
noahml
AI & ML interests
None yet
Recent Activity
commentedon a paper less than a minute ago
Why Far Looks Up: Probing Spatial Representation in Vision-Language Models commentedon a paper about 7 hours ago
When Cloud Agents Meet Device Agents: Lessons from Hybrid Multi-Agent Systems commentedon a paper about 7 hours ago
REPOT: Recoverable Program-of-Thought via Checkpoint RepairOrganizations
None yet
commented a
paper less than a minute ago
commented 5
papers about 7 hours ago
When Cloud Agents Meet Device Agents: Lessons from Hybrid Multi-Agent Systems
Paper • 2605.30102 • Published • 11 •
3
REPOT: Recoverable Program-of-Thought via Checkpoint Repair
Paper • 2605.30052 • Published • 6 •
3
Thinking Before Constraining: A Unified Decoding Framework for Large Language Models
Paper • 2601.07525 • Published • 6 •
3
Alignment Tampering: How Reinforcement Learning from Human Feedback Is Exploited to Optimize Misaligned Biases
Paper • 2605.27355 • Published • 2 •
3
Convex Low-resource Accent-Robust Language Detection in Speech Recognition
Paper • 2605.23235 • Published • 3 •
3
commented 3
papers 1 day ago
LoGeR: Long-Context Geometric Reconstruction with Hybrid Memory
Paper • 2603.03269 • Published • 63 •
7
Got a Secret? LLM Agents Can't Keep It: Evaluating Privacy in Multi-Agent Systems
Paper • 2605.27766 • Published • 1 •
3
EarlyTom: Early Token Compression Completes Fast Video Understanding
Paper • 2605.30010 • Published • 27 •
3
commented 11
papers 2 days ago
Can Vision-Language Models Solve the Shell Game?
Paper • 2603.08436 • Published • 39 •
4
SAHOO: Safeguarded Alignment for High-Order Optimization Objectives in Recursive Self-Improvement
Paper • 2603.06333 • Published • 1 •
3
Small Vision-Language Models are Smart Compressors for Long Video Understanding
Paper • 2604.08120 • Published • 20 •
3
Meta-learning In-Context Enables Training-Free Cross Subject Brain Decoding
Paper • 2604.08537 • Published • 9 •
3
Time is Not a Label: Continuous Phase Rotation for Temporal Knowledge Graphs and Agentic Memory
Paper • 2604.11544 • Published • 4 •
3
Models That Know How Evaluations Are Designed Score Safer
Paper • 2605.28591 • Published • 6 •
5
SWE-AGILE: A Software Agent Framework for Efficiently Managing Dynamic Reasoning Context
Paper • 2604.11716 • Published • 5 •
3
ProRL: Effective Reinforcement Learning for Proactive Recommendation via Rectified Policy Gradient Estimation
Paper • 2605.28293 • Published • 81 •
3
Models That Know How Evaluations Are Designed Score Safer
Paper • 2605.28591 • Published • 6 •
5
Self-Improving Language Models with Bidirectional Evolutionary Search
Paper • 2605.28814 • Published • 54 •
3
CollectionLoRA: Collecting 50 Effects in 1 LoRA via Multi-Teacher On-Policy Distillation
Paper • 2605.25378 • Published • 53 •
3