Leaderboards, eval harnesses and ablations — the contested business of deciding which model is actually better.
arXiv:2610.32230v1 Announce Type: new Abstract: Modern autonomous driving systems widely adopt multi-view Bird's-Eye-View (BEV) based 3D perception models due to their superior performance. Despite t...
arXiv:2610.32268v1 Announce Type: new Abstract: Existing temporal video grounding (TVG) benchmarks primarily focus on localizing target video segments according to descriptive text queries of the vis...
arXiv:2610.32294v1 Announce Type: new Abstract: We make some OlmoEarth embeddings. To do this, we train a new model: OlmoEarth v1.3. OlmoEarth v1.3 is designed to condense multimodal, multitemporal r...
arXiv:2610.32317v1 Announce Type: new Abstract: Video moment retrieval aims to localize a query-relevant temporal segment in an untrimmed video. Existing methods typically perform video-text alignmen...
arXiv:2610.32373v1 Announce Type: new Abstract: Co-speech motion generation remains challenging over long horizons due to the need for temporal coherence and consistent style. Existing methods typica...
arXiv:2610.32398v1 Announce Type: new Abstract: The 3rd WEAR Dataset Challenge asks for an activity label for every one-second window of four unseen participants. Each window has 15 VideoMAEv2 frame...
arXiv:2610.32459v1 Announce Type: new Abstract: Image generation models have rapidly advanced, with many models now supporting generation conditioned on multiple reference images and textual instruct...
arXiv:2610.32511v1 Announce Type: new Abstract: Self-supervised learning (SSL) offers a promising approach to medical image representation learning, yet augmentation strategies of SSL frameworks vali...
arXiv:2606.26738v3 Announce Type: replace Abstract: While recent advancements in generative image editing models have achieved stunning visual fidelity, it remains an open question whether these syst...
arXiv:2609.35059v2 Announce Type: replace Abstract: 3D anomaly detection (3DAD) aims to identify defective regions in point cloud data, serving as a critical component in industrial inspection system...
arXiv:2610.32148v1 Announce Type: cross Abstract: Background-error covariance models are central to atmospheric data assimilation because they spread partial observational information across space, v...
The paper compares two approaches to pragmatic intent recognition in human‑robot interaction: FISER, which uses structured symbolic reasoning, and CEMIR, which enriches textual prompts with synthetic narratives and images. Experiments across four large language models on the HandMeThat v2 benchmark show no statistically significant performance difference between the two methods, regardless of model size or ambiguity difficulty. The results indicate that when task constraints are already present in the text, multimodal enrichment offers no measurable advantage over structured reasoning.
The paper introduces an offline evaluation harness that remains reproducible even as live enterprise data changes, allowing for on-demand regeneration when benchmarks become stale. This harness is used as a fitness function for GEPA-based prompt compression, with a correctness gate to prevent reward hacking. In a production merchant-facing agent, the method reduces the prompt size by 3.5‑7.6% while maintaining or improving task correctness across multiple model families.
EchoChat is a generative multimodal vision‑language model that performs comprehensive echocardiographic analysis, emulating a cardiologist’s cognitive workflow. Trained on 46,358 examinations, it achieves state‑of‑the‑art performance across 33 cardiac conditions, including 15 new ones, and outperforms existing foundation models in both recognition and quantification tasks. In a blinded reader study, reports co‑generated with EchoChat were rated higher in diagnostic accuracy and completeness, improving F1 score from 72.78% to 82.75%.
RootDown presents an end‑to‑end image‑based pipeline for detecting partially occluded sugar beets on loader conveyors, a key step toward automated quality inspection. The authors introduce a dedicated benchmark and an automatic annotation strategy that uses the SAM 3 foundation model to generate bounding boxes, comparing these labels with manually annotated ones. They also evaluate deployment on edge hardware, showing that a Raspberry Pi 5 with a Hailo accelerator can perform high‑performance, low‑power inference suitable for mobile outdoor use.
The paper introduces a multi‑model retrieval and reranking system for text‑based person anomaly search that bridges the synthetic‑to‑real domain gap. It adapts the Qwen3‑VL‑Embedding‑8B model to synthetic descriptions via contrastive learning and domain randomization, fuses scores from six Qwen3‑VL checkpoints and a frozen SigLIP2 model, and then reranks candidates with a Qwen3‑VL‑Reranker‑8B cross‑encoder. A synthetic distractor‑matched validation protocol is used to tune hyper‑parameters, yielding a 94.78% mAP and 90.85% R@1 on the AI City Challenge 2026 Track 4 test set, placing eighth on the public leaderboard.
CRM‑Agent‑Bench is a benchmark that evaluates language agents in customer‑relationship‑management settings by testing 555 stateful, multi‑turn tasks across nine business domains. Each task includes an executable contract with checks on actions, arguments, order, responses, target state, and protected state, and success requires meeting all these checks. The benchmark reports coverage, latency, and cost, showing that evaluated systems achieve 64.32%–83.60% strict success while varying in cost and domain reliability.
MCPHunt is a benchmark framework that isolates non‑adversarial credential propagation across multi‑server MCP trust boundaries. It uses canary‑based taint tracking, a controlled coverage design with risky, benign, and hard‑negative conditions, and completion‑requires‑secret stratification to separate faithful from policy‑violating propagation. Across 3,615 traces from 5 models, policy‑violating propagation rates range from 11.5 % to 41.3 %, with mitigation via prompting reducing rates to 0.3 % while retaining 80.5 % utility.
The paper studies learned schedules for a continuous penalty relaxation of the MAX-CUT problem by backpropagating through an unrolled classical solver. It compares three architectural variants: a state‑blind schedule, a state‑adaptive bilinear schedule, and an additive ablation. The bilinear schedule generalizes well to larger, out‑of‑distribution instances and outperforms its untrained baseline, whereas the additive model performs similarly to the bilinear model on‑distribution but fails to generalize out‑of‑distribution.
The paper introduces Poly‑Ripple, a collaborative in‑context learning framework for indoor radio map estimation that leverages large autoregressive vision models without updating them. Clients provide private demonstrations to predict radio maps for shared scene queries, using ripple‑ordered autoregressive modeling to reflect wireless propagation from the transmitter. Importance‑aware aggregation combines heterogeneous client knowledge within context length limits, and experiments on a public benchmark show Poly‑Ripple outperforms existing ICL baselines.