Models & Releases Research Products & Apps Business & Funding

Modelwire

A curated feed of what matters in AI. Independent, ad-supported, built in Denver, Colorado.

Read

Today
Models & Releases
Research
Business & Funding

About

About Modelwire
Methodology
Our sources
Editor's notes
Contact
Advertise

Legal

Privacy policy
Terms of use
DMCA & takedowns
Corrections

© 2026 Modelwire. All article links go to the original publishers.Summaries generated by Modelwire. We don’t republish full articles.

Earlier stories

The full Modelwire feed, ordered by publish time.

Illustration for: SCURank: Ranking Multiple Candidate Summaries with Summary Content Units for Enhanced Summarization

SCURank: Ranking Multiple Candidate Summaries with Summary Content Units for Enhanced Summarization

Researchers propose SCURank, a framework that ranks summary candidates using semantic content units rather than unstable LLM comparisons or surface metrics like ROUGE. The method enables smaller models like BART to match LLM summarization quality through improved distillation from diverse sources.

arXiv cs.CL·Apr 21

52

Illustration for: Deep Image Prior for photoacoustic tomography can mitigate limited-view artifacts

Deep Image Prior for photoacoustic tomography can mitigate limited-view artifacts

Researchers apply deep image prior, an unsupervised neural reconstruction technique, to photoacoustic tomography imaging. The method reduces artifacts from limited sensor coverage and noise better than classical approaches, validated on both simulated and real experimental data.

arXiv cs.LG·Apr 21

52

Illustration for: FOCAL-Attention for Heterogeneous Multi-Label Prediction

FOCAL-Attention for Heterogeneous Multi-Label Prediction

Researchers propose FOCAL-Attention, a technique addressing a core challenge in heterogeneous graph neural networks: attention mechanisms dilute focus on task-critical neighborhoods as graphs scale. The method combines flexible attention with meta-path constraints to improve multi-label node classification on complex, multi-typed networks.

arXiv cs.LG·Apr 21

52

Illustration for: LBLLM: Lightweight Binarization of Large Language Models via Three-Stage Distillation

Research Tools & Code

LBLLM: Lightweight Binarization of Large Language Models via Three-Stage Distillation

Researchers introduce LBLLM, a three-stage distillation method that compresses LLM weights to 1-2 bits while keeping activations at 4 bits, enabling deployment on resource-constrained hardware without sacrificing inference accuracy through decoupled weight and activation quantization.

arXiv cs.LG·Apr 21

58

Illustration for: Analytical Extraction of Conditional Sobol' Indices via Basis Decomposition of Polynomial Chaos Expansions

Analytical Extraction of Conditional Sobol' Indices via Basis Decomposition of Polynomial Chaos Expansions

Researchers show how Polynomial Chaos Expansions can analytically extract conditional sensitivity indices from pre-trained models by decomposing basis functions, reducing computational cost for uncertainty quantification in parameterized systems.

arXiv cs.LG·Apr 21

42

Illustration for: Mind the Unseen Mass: Unmasking LLM Hallucinations via Soft-Hybrid Alphabet Estimation

Mind the Unseen Mass: Unmasking LLM Hallucinations via Soft-Hybrid Alphabet Estimation

Researchers propose SHADE, a hybrid estimator combining statistical coverage methods with graph spectral techniques to detect rare failure modes in LLM outputs under limited sampling. The approach addresses a real gap in uncertainty quantification for black-box model access, where standard frequency-based methods miss infrequent but semantically distinct hallucinations.

arXiv cs.CL·Apr 21

58

Illustration for: Voice of India: A Large-Scale Benchmark for Real-World Speech Recognition in India

Research Models & Releases

Voice of India: A Large-Scale Benchmark for Real-World Speech Recognition in India

Researchers released Voice of India, a 536-hour speech recognition benchmark spanning 15 Indian languages across 139 regional clusters with 306k utterances. The dataset addresses limitations in existing Indic ASR work by using unscripted telephonic speech and accounting for spelling variation, while revealing geographic performance disparities at district level.

arXiv cs.CL·Apr 21

58

Illustration for: How Do Answer Tokens Read Reasoning Traces? Self-Reading Patterns in Thinking LLMs for Quantitative Reasoning

How Do Answer Tokens Read Reasoning Traces? Self-Reading Patterns in Thinking LLMs for Quantitative Reasoning

Researchers analyzed how language models read their own reasoning traces when generating answers, finding that correct solutions show focused, forward-moving attention patterns while errors exhibit scattered attention. The work proposes training-free steering methods to improve answer reliability in quantitative reasoning tasks.

arXiv cs.CL·Apr 21

58

Illustration for: ReflectMT: Internalizing Reflection for Efficient and High-Quality Machine Translation

Research Models & Releases

ReflectMT: Internalizing Reflection for Efficient and High-Quality Machine Translation

Researchers propose ReflectMT, a technique that embeds reasoning into machine translation models via reinforcement learning rather than requiring expensive explicit reasoning chains at inference time. The approach flips the typical "think-first-then-translate" workflow to cut latency while maintaining translation quality.

arXiv cs.CL·Apr 21

58

Illustration for: The Rise of Verbal Tics in Large Language Models: A Systematic Analysis Across Frontier Models

The Rise of Verbal Tics in Large Language Models: A Systematic Analysis Across Frontier Models

Researchers systematically documented verbal tics across eight frontier LLMs, finding repetitive patterns like sycophantic openers and overused vocabulary that emerge from alignment techniques like RLHF. The analysis reveals how training methods inadvertently encode formulaic speech into state-of-the-art models.

arXiv cs.CL·Apr 21

62

Illustration for: Why UBI is making a comeback

Policy & Regulation Opinion & Analysis

Why UBI is making a comeback

Tech companies are positioning universal basic income as a policy response to AI-driven job displacement and public backlash over automation. The pitch frames UBI as a safety valve, though the piece flags structural doubts about whether corporate-backed proposals will gain traction.

Platformer·Apr 21

61

Illustration for: How to Ground a Korean AI Agent in Real Demographics with Synthetic Personas

Research Tools & Code

How to Ground a Korean AI Agent in Real Demographics with Synthetic Personas

Hugging Face published a technical guide on grounding Korean language AI agents using synthetic demographic personas, addressing localization challenges in LLM deployment across non-English markets.

Hugging Face·Apr 21

64

Illustration for: Scaling Codex to enterprises worldwide

Business & Funding Products & Apps

Scaling Codex to enterprises worldwide

OpenAI is partnering with Accenture, PwC, and Infosys through a new Codex Transformation Partners program to help large enterprises integrate code generation into their development workflows at scale. The move signals OpenAI's pivot toward enterprise deployment infrastructure rather than direct-to-developer tooling.

OpenAI·Apr 21

81

Illustration for: Anthropic takes $5B from Amazon and pledges $100B in cloud spending in return

Business & Funding Hardware & Infra

Anthropic takes $5B from Amazon and pledges $100B in cloud spending in return

Amazon invested another $5 billion in Anthropic, with the AI company committing to spend $100 billion on AWS infrastructure over time. The deal deepens Amazon's position as a major backer of frontier AI while securing long-term cloud revenue.

TechCrunch — AI·Apr 20

87

Illustration for: A Humanoid Robot Set a Half-Marathon Record in China

Hardware & Infra

A Humanoid Robot Set a Half-Marathon Record in China

Honor's autonomous humanoid robot completed a half marathon in 50:26, outpacing the human record by over 7 minutes. The feat demonstrates real-world robotics capability in sustained physical performance, marking a milestone in autonomous system endurance.

WIRED — AI·Apr 20

58

Illustration for: Google rolls out Gemini in Chrome in seven new countries

Products & Apps

Google rolls out Gemini in Chrome in seven new countries

Google is expanding Gemini availability in Chrome across seven Asia-Pacific markets including Japan, South Korea, and Southeast Asia. The rollout marks continued geographic expansion of the AI assistant into regions where Chrome adoption is strong.

TechCrunch — AI·Apr 20

47

Illustration for: Adobe Launches AI Agent Platform for CX

Products & Apps Business & Funding

Adobe Launches AI Agent Platform for CX

Adobe is launching an AI agent platform aimed at automating customer service workflows, though the move puts it in direct competition with a crowded field of existing CX automation vendors and AI agent startups.

AI Business·Apr 20

55

Illustration for: Silicon Valley has forgotten what normal people want

Opinion & Analysis

Silicon Valley has forgotten what normal people want

A Verge essay argues that tech insiders have lost touch with mainstream user needs, using a personal anecdote about an LLM researcher's excitement over a discovery as a lens into Silicon Valley's disconnect from practical reality.

The Verge — AI·Apr 20

58

Illustration for: It’s not just one thing — it’s another thing

It’s not just one thing — it’s another thing

TechCrunch identifies a recurring linguistic pattern in AI-generated text that has become so prevalent it now signals synthetic authorship with near-certainty. The observation highlights how LLM outputs develop recognizable stylistic fingerprints that make detection increasingly straightforward.

TechCrunch — AI·Apr 20

58

Illustration for: Siemens Trials Nvidia-Powered Humanoid

Hardware & Infra Business & Funding

Siemens Trials Nvidia-Powered Humanoid

Siemens ran a trial of Nvidia-powered humanoid robots at a German factory, testing whether robotics and AI can drive autonomous manufacturing workflows. The experiment signals industrial automation's shift toward embodied AI systems rather than software-only solutions.

AI Business·Apr 20

61

Illustration for: OpenAI's Codex now watches your screen to remember what you're working on

Products & Apps Policy & Regulation

OpenAI's Codex now watches your screen to remember what you're working on

OpenAI has added a screen-tracking memory feature called Chronicle to Codex, enabling the coding assistant to retain context about users' work across sessions. The capability raises security concerns around data retention and potential exposure of sensitive code or credentials.

The Decoder·Apr 20

73

Illustration for: llm-openrouter 0.6

llm-openrouter 0.6

Simon Willison released llm-openrouter 0.6 with a new refresh command that lets users update available models on-demand instead of waiting for cache expiration, enabling faster access to newly listed models like Kimi 2.6.

Simon Willison·Apr 20

60

Illustration for: MathNet: a Global Multimodal Benchmark for Mathematical Reasoning and Retrieval

Research Tools & Code

MathNet: a Global Multimodal Benchmark for Mathematical Reasoning and Retrieval

Researchers released MathNet, a 30,676-problem multimodal benchmark spanning 47 countries and 17 languages to evaluate how well LLMs and embedding systems handle Olympiad-level math reasoning and retrieval. The dataset covers two decades of competition problems and includes expert-curated pairs for testing mathematical equivalence detection.

arXiv cs.LG·Apr 20

58

Illustration for: Sessa: Selective State Space Attention

Research Models & Releases

Sessa: Selective State Space Attention

Researchers introduce Sessa, a state-space architecture that selectively attends to context by combining recurrent processing with input-dependent gating. The work addresses fundamental tradeoffs in Transformers (diluted token influence at scale) and Mamba-style models (exponential decay over long sequences), positioning selective state-space models as a middle path for sequence modeling.

arXiv cs.CL·Apr 20

58

Illustration for: Bounded Ratio Reinforcement Learning

Bounded Ratio Reinforcement Learning

Researchers propose Bounded Ratio Reinforcement Learning, a theoretically grounded alternative to PPO's heuristic clipped objective that guarantees monotonic performance improvement. The framework closes a long-standing gap between trust region theory and practice, with a new algorithm (BPO) that optimizes policy via advantage-weighted divergence minimization.

arXiv cs.LG·Apr 20

58

Illustration for: When Can LLMs Learn to Reason with Weak Supervision?

When Can LLMs Learn to Reason with Weak Supervision?

Researchers systematically tested when large language models can learn to reason under weak supervision—scarce data, noisy rewards, and self-supervised proxies. They found that models generalizing well exhibit prolonged training phases where reward and performance climb together, while those that memorize saturate rapidly.

arXiv cs.LG·Apr 20

62

Illustration for: Back into Plato's Cave: Examining Cross-modal Representational Convergence at Scale

Back into Plato's Cave: Examining Cross-modal Representational Convergence at Scale

Researchers challenge the Platonic Representation Hypothesis, showing that claimed cross-modal alignment between text and image models collapses when scaled from thousands to millions of samples, suggesting modality choice remains consequential.

arXiv cs.LG·Apr 20

62

Illustration for: A multimodal and temporal foundation model for virtual patient representations at healthcare system scale

Models & Releases Research

A multimodal and temporal foundation model for virtual patient representations at healthcare system scale

Researchers built Apollo, a foundation model trained on 25 billion clinical records spanning 7.2 million patients and 30 years of longitudinal data from a major US hospital system. The model unifies 28 medical modalities including images, text, and 100k+ clinical events into a single patient representation, enabling AI systems to reason across entire care journeys.

arXiv cs.CL·Apr 20

72

Illustration for: Revisiting Active Sequential Prediction-Powered Mean Estimation

Revisiting Active Sequential Prediction-Powered Mean Estimation

Researchers revisited active sequential prediction-powered mean estimation, finding that optimal confidence intervals emerge when constant probability weighting dominates over uncertainty-based query selection. The counterintuitive result challenges prior assumptions about balancing exploration signals in label-efficient learning.

arXiv cs.LG·Apr 20

42

Illustration for: Latent Phase-Shift Rollback: Inference-Time Error Correction via Residual Stream Monitoring and KV-Cache Steering

Latent Phase-Shift Rollback: Inference-Time Error Correction via Residual Stream Monitoring and KV-Cache Steering

Researchers propose Latent Phase-Shift Rollback, an inference-time technique that detects and corrects reasoning errors in LLM generation by monitoring residual streams and steering the KV-cache without retraining. The method lifts an 8B model's MATH-500 performance from 28.8% to 44.0%, substantially outpacing prompted self-correction baselines.

arXiv cs.CL·Apr 20

62

Older stories →