Method to distill Answer-Set Programming (ASP) rules from LLMs for Visual Question Answering (VQA). The approach uses VQA dataset examples to guide the LLM in extending an initial reasoning theory, with validation by the ASP solver. Demonstrates effectiveness across multiple datasets with few examples.
Comparative study of brain region importance (EEG) for cognitive workload prediction. Analysis across 4 public datasets: frontal electrodes outperform full-scalp baseline by 15-20% in relative rank position using fewer sensors. Fronto-central regions show most stable predictive utility.
Padyam2Gadyam: dataset of 600 classical Telugu poems (13th-17th centuries) with human-verified modern Telugu and English prose translations. Evaluation of 5 contemporary LLMs shows insufficient performance and limitations of existing MT evaluation metrics for this task.
Internal Coherence Maximization (ICM) automatically generates examples to align AI models with diverse human values without extensive human supervision. Tested across four benchmarks, ICM matches gold label performance. Example coherence improves generalization even at constant accuracy, especially for underrepresented personas.
SNMPBB, first adaptation of nonmonotone Barzilai-Borwein methods to Symmetric NMF. Demonstrates 6× speedup over SymANLS on synthetic data. Extensions for graph clustering (Graph-SNMPBB) and large-scale problems (LAI-SNMPBB) with proven global convergence to first-order stationary points.
MASC (Margin Self-Correction) is a language-model unlearning method that efficiently reduces the logit gap between the original next token and its alternatives, without requiring downstream evaluation. Tested on TOFU, MUSE News, and MUSE Books, it achieves competitive forget-retain trade-offs at a fraction of existing baselines' computational cost.
Comparative study of encoder-only Transformer vs LSTM for streamflow prediction in ungauged basins. On NOAA NWM data, LSTM outperforms Transformer. Incorporating downstream information improves all models (+60% median NNSE). Recurrent memory proves better suited for this hydrologic inference task.
arXiv study showing LLMs struggle to infer user sociodemographics from single conversation histories. Disparities in advice (legal, medical, financial) are minimal but present. Conversation topics prove more predictive than sociodemographic data and affect LLM outputs unpredictably.
ChatHealthAI aligns structured EHR representations from a pretrained EHR foundation model with a frozen LLM's semantic space via a task-aware resampler. The multimodal framework integrates longitudinal patient representations with refined clinical event descriptions, improving interpretable clinical reasoning while maintaining competitive predictive performance on the EHRSHOT benchmark.
Empirical study detecting natural experiments (implicit interventions) in real-world datasets using causal discovery and feature selection. Authors validate on synthetic data then evaluate 50+ real datasets, showing that treating data as interventional rather than observational improves model performance.
arXiv paper on LLM representations: lexical overlap influences embeddings more than semantic content, persists across layers and architectures, and degrades performance in summarization and model editing tasks.
Deterministic orchestration framework for validating fragmented ESG data (Scope 1-3) with temporal anomaly detection, imbalance-aware ensemble learning, and audit provenance tracing. Synthetic benchmark calibrated against GHG Protocol, PCAF, ISSB standards. Evaluation on classification, calibration, and provenance chain completeness metrics.
Theoretical paper on generalization bounds under distribution shift. Proposes framework quantifying extra risk from regime composition mismatch in Markov-switching environments, with exact decomposition and minimax lower bounds. Tested on 25 years of global equity indices.
Adaptive multifidelity algorithm for machine learning in quantum chemistry. Dynamically determines dataset composition by querying samples at each fidelity level. Reduces data generation costs by up to 30× versus single-fidelity methods and 5× versus standard MFML on coupled cluster energies and excitation energies.
SpecFlow introduces a lightweight multimodal spatial reasoning framework representing intermediate visual thoughts in fixed-size discrete cosine space. Classifier-free guidance enables autoregressive textual thoughts to steer visual state updates without context expansion. Achieves competitive reasoning performance while reducing computation and KV cache costs by up to 2.1×.
DMT-CBT models longitudinal therapeutic state evolution in CBT across multiple sessions. The framework maintains structured states, incorporates multimodal data and tool-augmented interventions. DMTCorpus, a synthetic multimodal dataset, demonstrates improved counseling fidelity and therapeutic alliance over post-hoc extraction approaches.
MiniMax introduces a new attention architecture (MSA) natively supporting 1M tokens without quadratic complexity. 'KV outer gather Q' approach delivers 4× faster than Flash-Sparse-Attention, compute reduced to 1/20th, 9× prefilling speedup, 15× decoding speedup. First open-weight model combining frontier coding, 1M context, and native multimodality.
Release of datasette-agent-micropython 0.1a0. Enables Datasette Agent to safely generate and execute Python code via WebAssembly sandbox. GPT-5.5 failed to break out of the sandbox in testing.
Since June 2026, European digital authorities using Qwen or DeepSeek must comply with the AI Act requiring disclosure of training data. Beijing refuses to share them, creating a major regulatory conflict between the EU and Chinese providers.
Microsoft releases MAI-Code-1-Flash, a 5B active parameter model achieving 51% on SWE-Bench Pro. This compact architecture demonstrates competitive performance on complex software engineering tasks.
OpenAI expands Codex with role-specific plugins for data analysis, sales, and investment banking. 5 million weekly users, 20% non-developers. This segment grows 3x faster than developers.
Gemma 4 E4B in Google's LiteRT format achieves 157.2 tok/s text generation, 2.4× faster than Q4 GGUF (66.3 tok/s) via multi-token prediction (MTP). Image captioning shows only 1.1× speedup as vision encoder is the bottleneck. Tested on RTX 4060 Ti 16GB.
KeyLM, a 75M parameter model trained on 18B tokens, outperforms SmolLM-135M-Instruct on IFEval (17.85 vs 17.15) despite half the size and 30x less training data. Standard architecture: GQA, RoPE, SwiGLU, 24 layers, trained on FineWeb-Edu, Wikipedia, Reddit and public datasets.
Benchmark of 20 small LLMs on RTX 4050 6GB GPU. Author tests Q4/Q6 GGUF quantizations with 6 qualitative probes (tool-call, strict JSON, plan decomposition, no path hallucination) instead of full suites, measuring prefill speed and generation at 1k/8k/32k tokens to identify viable models for local inference on constrained hardware.
The Linux Foundation unveils DNS-AID (Agent IA Discovery), an open-source project using DNS as a discovery layer for AI agents. This decentralized approach challenges proprietary registries from hyperscalers (OpenAI, Google, Anthropic).
Context-mode optimizes context window for AI coding agents by sandboxing tool outputs. Achieves 98% token reduction. Compatible with 15 platforms.
Headroom compresses tool outputs, logs, files, and RAG chunks before sending to LLM. Reduces token consumption by 60-95% without degrading answers. Available as library, proxy, and MCP server.
2-week trial: replaced Claude with Qwen3.6-27B (Q6_K, 22GB) in multi-agent orchestrator on RTX 3090. Wins: plan generation (95% valid), memory extraction, auto-review. Failures: tool-call reliability 12% error rate vs 0.5% Claude, context drift >14k tokens, weak cascade-failure handling.
Qwen 3.6-35B-A3B achieves 977 tokens/s prompt processing and 262k context window on Intel Arc B70 Pro via llama.cpp with SYCL backend. User reports stable, usable local inference for complex tasks including game generation.
Niels from Hugging Face announces paperswithcode.co, a revived SOTA tracking platform. New feature: indexing major conferences (NeurIPS, CVPR, ICML). CVPR 2026 papers indexed with arXiv IDs, categorized by task, tagged with GitHub, project pages, Hugging Face artifacts, and evals.
Online, distribution-free framework for controlling Conditional Value-at-Risk (CVaR) in non-stationary and adversarial environments. Combines conformal tail risk control, online learning, and Rockafellar-Uryasev variational representation. Provable safety guarantees for nonlinear tail risk under arbitrary data-generating processes. Applications: portfolio risk management and LLM toxicity mitigation.
Empirical study adapting Phi Silica for short-form text rewriting via dataset curation, prompt distillation, and parameter-efficient fine-tuning. Dataset built from public slide decks. Results: improved semantic fidelity, reduced hallucinations, higher preference win rate vs GPT-5-chat.
Academic paper proposing product-aware autoencoders for anomaly detection in multi-product cyber-physical systems. Traditional global models create blind spots where attacks can evade detection. Tests on Tennessee Eastman Process benchmark: product-aware model achieves 100% detection accuracy versus 22.2% for global baseline in attack scenarios.
Study of LLM agent faithfulness in a Texas Poker simulator. Researchers measure the gap between stated reasoning and actual actions by decomposing the problem into two steps: reasoning-conclusion and conclusion-action. The two steps exhibit opposite behaviors.
New automated metric (Triangulated Preference Shift score) to measure lexical biases introduced during preference learning (RLHF) in LLMs without manual curation. Analysis across 6 model families reveals a shift toward a 'language of prestige' (overuse of 'delve', 'furthermore').
A legal framework for agentic AI tort liability. The paper proposes three interaction types (autonomous drift, pure tool use, collaborative planning) and uses interaction logs as primary evidence to determine where liability attaches. Introduces a "Reasonable Agent" standard based on constraint verification and forensic logging.
Study on adaptive data selection for wearable prediction across 3 modalities (heart rate, activity, EMA) in longitudinal dataset. AUROC gains up to 0.7 for low-baseline participants, strong inverse correlation (r=-0.67) between improvement and baseline performance. Limited benefit for strong baselines.
arXiv paper on temperature's role in LLM distillation. Authors show forward KL (FKL) outperforms reverse KL (RKL) at higher temperatures, contrary to prior empirical conclusions that omitted this parameter. Temperature enriches FKL with non-dominant token signals while merely rescaling RKL gradients.
New quantization method preserving inner products with unseen vectors. Adaptive unbiased algorithms developed with theoretical guarantees. Practical implementations 2-10× faster than prior ASQ state-of-the-art while maintaining quality.
ReMax formalizes exploration in RL through retries: a policy is evaluated by expected maximum return over M samples. Exploration emerges naturally without explicit bonuses. RePPO, a PPO variant optimizing ReMax, generalizes discrete M to continuous parameter m for fine-grained exploration control. Results on MinAtar and Craftax benchmarks.