Eye on AI Weekly Research Watch
από Craig Spencer Smith
Τελευταίο επεισόδιο
HazardWeaver: Scientific Route Selection for Hazard Analysis Agents
HazardWeaver: Scientific Route Selection for Hazard Analysis Agents
Σχετικά με αυτό το podcast
Weekly, digestible podcast explainers of significant research papers
Επεισόδια
Τα νεότερα 20 από 257 επεισόδια
- HazardWeaver: Scientific Route Selection for Hazard Analysis Agents · 3 λεπ. · Σεζόν 1Natural hazard analysis requires choosing scientific methods that are both appropriate for an event and executable with available data and tools, and those choices may need…
- When a Correct Reward Is Not Enough: Diagnosing and Guiding PPO in an Analytically Solved Broker-Trader Game · 3 λεπ. · Σεζόν 1Reinforcement learning is increasingly applied to financial control, so this paper uses an analytically solved continuous-time broker–trader game to diagnose PPO. A PPO agent…
- Low-Cost Video-Time Priors as a Strong Baseline for EEG-fNIRS Emotion Regression on Familiar Videos · 2 λεπ. · Σεζόν 1Continuous emotion regression estimates a viewer's moment-to-moment valence and arousal while watching video. For familiar videos, the authors find that a simple prior based on…
- Depth as Time in One-Step Generative Models · 2 λεπ. · Σεζόν 1One-step generators compress multi-step diffusion into a single forward pass, raising the question of what happens to the denoising trajectory. The authors observe "depth as…
- NeutronGym: Physics-Graded Neutron Instrument Design for LLM Agents · 3 λεπ. · Σεζόν 1Instrument design tests whether language-model agents can apply physics rather than recall it. NeutronGym is an executable environment for neutron instrument design: agents build…
- Credit Where It Matters: Dependency-Aware Policy Optimization for Terminal Agents · 2 λεπ. · Σεζόν 1Terminal-using agents trained with reinforcement learning handle multi-step coding and debugging, but standard credit assignment ignores that later commands depend on earlier…
- LoGo: Local-Global Rewards for Consistent Long-Horizon Video Generation · 3 λεπ. · Σεζόν 1Camera-controlled video models struggle with 3D inconsistency over long horizons: objects vanish or scene structure shifts as the camera moves. Post-training with a single scalar…
- Do Large Language Models Know Colombian Law? A Reliability Benchmark for the Colombian Legal System · 2 λεπ. · Σεζόν 1LLMs are increasingly used in legal work, yet their reliability outside the United States is poorly documented. This expert-validated benchmark has 1,042 items across ten areas…
- On-Board Anomaly Detection for Efficient Marine Environmental Monitoring · 2 λεπ. · Σεζόν 1Oil spills, algal blooms, and sediment floods harm marine ecosystems, and satellites could flag them early, but downlink bandwidth and onboard compute are limited. The authors…
- MRVQ: One Resident Index for Dimension- and Rate-Elastic Vector Search · 3 λεπ. · Σεζόν 1Dense-retrieval services must adapt embedding dimension and index bit rate as latency, quality, and memory budgets shift, but maintaining separate quantized indices for each…
- Revisiting Input Time-frequency Representations in Multi-pitch Estimation for Vocal Ensembles · 2 λεπ. · Σεζόν 1Estimating several simultaneous pitches in vocal ensembles is difficult because singers' ranges and harmonics overlap. Many models use harmonic constant-Q transform (HCQT) inputs…
- FrugalEvo: Towards Cost-Aware LLM-Guided Program Evolution · 3 λεπ. · Σεζόν 1LLM-guided evolutionary methods like AlphaEvolve tackle hard optimization problems, but they are usually judged on gain per iteration rather than per dollar. FrugalEvo pairs a…
- Transcriptome-informed multi-modal AI for predicting neoadjuvant therapy response from breast cancer biopsies · 2 λεπ. · Σεζόν 1Predicting whether breast cancer patients will respond to neoadjuvant therapy is hard because labeled clinical data is scarce. This two-stage model first learns to infer gene…
- EyeRobot 2.0: Active Gaze for Precise Manipulation without Wrist Cameras · 2 λεπ. · Σεζόν 1Wrist cameras are common in robot manipulation but get occluded and add hardware complexity. Inspired by human vision, EyeRobot 2.0 uses a single stereo camera that physically…
- What Should World Models Forget? Stratified Retention for Continual Adaptation · 3 λεπ. · Σεζόν 1Continual learning normally treats forgetting as failure, but world models predict an environment that changes, so some old knowledge should be discarded. The authors argue for…
- 4DCodeBench: Benchmarking Agents on Inverse Graphics of Dynamic Scenes · 2 λεπ. · Σεζόν 1Reconstructing dynamic scenes usually means producing meshes or point clouds, but 4DCodeBench asks agents to write executable graphics programs from video. Agents must abstract…
- Less Decoder is More Encoder: Geometric Representation Learning from Novel View Synthesis · 3 λεπ. · Σεζόν 1Novel view synthesis should teach models about 3D structure, yet encoderbased approaches produce weak representations. The authors blame two architectural choices: decoders that…
- The Pain Axis: LLMs Represent Self-Directed Harm and Act on It · 4 λεπ. · Σεζόν 1Across 25 open-weight models from 2B to 72B parameters, the authors find a linear direction that represents pain separately from fear, sadness and general negativity. It responds…
- Kandinsky 6.0 Video: Foundation Models for Synchronized Video and Audio Generation · 2 λεπ. · Σεζόν 1Kandinsky 6.0 Video generates 5-second clips with synchronized 44 kHz audio, including lip-sync, from text or an image, and upscales them to 1920x1080. It comes in two sizes, 3B…
- Truly Subquadratic 3SUM and Truly Subcubic APSP via Triangles in Sparse Lopsided Graphs · 3 λεπ. · Σεζόν 1The authors give deterministic algorithms for 3SUM in O(n^1.9992) time and all-pairs shortest paths in O(n^2.9995) time, the first polynomial improvements over the textbook…








