Machine Learning News

A study from arXiv (2608.00507v1) training a ~7M-parameter Transformer encoder on child-directed and read speech found that the learning objective, not architecture, determines perceptual narrowing in self-supervised spe…

Researchers propose the Capability-Driven Multimodal Scaling Law, the first cross-family framework that predicts vision-language model (VLM) benchmark accuracy from textual capability scores, trained on over 150 VLMs acr…

Researchers trained a small language model to route queries to specialized retrieval agents, achieving an NDCG@10 of 0.918 on mismatched queries, compared to 0.539 for Amazon Nova Lite and 0.490 for Claude Haiku 4.5, and…

A new arXiv preprint (2608.00585v1) shows that per-chunk verification fails on multi-hop retrieval-augmented generation, with entailment scoring reaching only 0.643, 0.523, and 0.560 AUC on HotpotQA, 2WikiMultihopQA, and…

Researchers propose SeDeM, a selective decompression framework that decouples compact memory storage from decoder conditioning for long-context question answering. On four long-context QA benchmarks, SeDeM achieves highe…

A new arXiv paper (2608.00582v1) introduces BPE-guided insertion, a method to adapt byte-level BPE tokenizers to underrepresented languages without changing the model vocabulary size. On Ukrainian adaptations of Nemotron…

Researchers at Observatorio Lázaro have built a self-populating database that has recorded more than two million English lexical borrowings (anglicisms) in the Spanish digital press, processing 1.88 million articles and …

A new arXiv preprint (2608.00078v1) presents a mobile deployment pipeline that uses QLoRA fine-tuning of an LLM to generate convolutional neural networks, with INT8 TensorFlow Lite export and on-device benchmarking on a …

A new arXiv preprint (2608.01014v1) introduces Cloud-ScPO, a topology-guided preference-mining framework that derives preference supervision from the hidden-state geometry of large language models (LLMs) for semi-supervi…

Researchers propose RSTG (Recovering Learning Signals via Adaptive Teacher Guidance), a method that selectively applies on-policy distillation to negative zero-variance prompts in reinforcement learning with verifiable r…

Researchers introduced OoO-Spec, an out-of-order semantic speculation method that speeds up LLM tool calling by predicting function choices and argument values in parallel. Across seven targets and three benchmarks, OoO-…

Snyk Secrets, an ML-powered secrets detection engine from Snyk's acquisition of BitPatrol, is now generally available as part of the Snyk AI Security Platform, aiming to reduce false positives and integrate into agentic …

A new arXiv preprint (2608.00246v1) proposes a retain-aware localization method for machine unlearning that considers parameter importance to both forgotten and retained data, reducing collateral damage to semantically s…

A new arXiv preprint (2608.00301v1) proves that error-penalized scoring rules, which reward correct answers with +1, penalize wrong ones with -λ, and give 0 for abstention, can cause KL-anchored gradient learners to drif…

Researchers introduced a flow-map learning (FML) framework to model unknown nonlocal partial differential equations (PDEs) directly from solution data, bypassing explicit evaluation of nonlocal operators. The method lear…

Researchers introduced DLLM-TTS, a text-to-speech framework using block discrete diffusion over X-Codec2 neural audio codec tokens, achieving a real-time factor of 0.15 with a 0.6B-parameter model trained on 20K hours, a…

Researchers at arXiv introduced a novel pre-training method for logographic character recognition that uses semantic-based contrastive learning to improve visual representations on imbalanced datasets. The method, which …

Researchers propose a multimodal machine-learning framework for calibration monitoring and quality assessment in archaeological digitisation, integrating photogrammetric 3D reconstruction, hyperspectral imaging, X-ray fl…

A new arXiv study (2608.00077v1) audits visual-token pruning in OCR-critical multimodal large language models (MLLMs), finding that accuracy alone misses failures where correct answers lack local token provenance. On loc…

A 115.7M-parameter masked diffusion language model trained with Explorative Modeling (XM) at K=3 for 10,000 steps on one RTX 4090 matched the compute of a 20,000-step baseline but showed a 0.19 nats higher test NELBO (4.…