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: Asset Harvester: Extracting 3D Assets from Autonomous Driving Logs for Simulation

Research Tools & Code

Asset Harvester: Extracting 3D Assets from Autonomous Driving Logs for Simulation

Researchers developed Asset Harvester, an image-to-3D pipeline that converts real driving footage into simulation-ready 3D object assets for autonomous vehicle testing. The system bridges a gap in closed-loop AV simulation by extracting complete, manipulable objects from sparse real-world observations rather than relying on single-model approaches.

arXiv cs.LG·Apr 20

58

Illustration for: An Integrated Deep-Learning Framework for Peptide-Protein Interaction Prediction and Target-Conditioned Peptide Generation with ConGA-PePPI and TC-PepGen

Research Models & Releases

An Integrated Deep-Learning Framework for Peptide-Protein Interaction Prediction and Target-Conditioned Peptide Generation with ConGA-PePPI and TC-PepGen

Researchers integrated two deep-learning models for peptide drug discovery: ConGA-PepPI predicts protein-binding sites with cross-attention mechanisms, while TC-PepGen generates candidate peptides conditioned on target proteins. The framework accelerates early-stage screening by combining prediction and generative capabilities in a single pipeline.

arXiv cs.LG·Apr 20

52

Illustration for: Semantic Step Prediction: Multi-Step Latent Forecasting in LLM Reasoning Trajectories via Step Sampling

Semantic Step Prediction: Multi-Step Latent Forecasting in LLM Reasoning Trajectories via Step Sampling

Researchers refined Semantic Tube Prediction by sampling at reasoning step boundaries rather than random token spans, achieving 168x better multi-step latent forecasting on ProcessBench versus 4x for the original approach. The technique improves how LLMs structure internal reasoning trajectories during fine-tuning.

arXiv cs.LG·Apr 20

52

Illustration for: ESsEN: Training Compact Discriminative Vision-Language Transformers in a Low-Resource Setting

ESsEN: Training Compact Discriminative Vision-Language Transformers in a Low-Resource Setting

Researchers demonstrate that two-tower encoder architectures outperform single-tower designs for vision-language tasks when training data and model size are constrained, drawing inspiration from how children learn language with sparse input.

arXiv cs.CL·Apr 20

52

Illustration for: NSA spies are reportedly using Anthropic’s Mythos, despite Pentagon feud

Policy & Regulation Business & Funding

NSA spies are reportedly using Anthropic’s Mythos, despite Pentagon feud

The NSA has reportedly deployed Anthropic's restricted Mythos model despite ongoing tensions between the intelligence agency and the Pentagon over AI governance. The move signals internal U.S. government fragmentation on which AI systems to operationalize for sensitive work.

TechCrunch — AI·Apr 20

69

Illustration for: 🔬 Training Transformers to solve 95% failure rate of Cancer Trials , Ron Alfa & Daniel Bear, Noetik

Business & Funding Products & Apps

🔬 Training Transformers to solve 95% failure rate of Cancer Trials , Ron Alfa & Daniel Bear, Noetik

Noetik's AI platform for patient-tumor matching just landed a $50M licensing deal with GSK, betting that better treatment stratification can flip the 95% cancer trial failure rate without discovering new drugs. The deal signals biotech's shift from AI-as-discovery to AI-as-platform, a structural change in how big pharma deploys ML.

Latent Space·Apr 20

85

Illustration for: BhashaSutra: A Task-Centric Unified Survey of Indian NLP Datasets, Corpora, and Resources

Research Tools & Code

BhashaSutra: A Task-Centric Unified Survey of Indian NLP Datasets, Corpora, and Resources

Researchers released BhashaSutra, the first comprehensive survey of Indian NLP resources, cataloging 200+ datasets, 50+ benchmarks, and 100+ models across 22 scheduled languages. The work addresses a gap in low-resource and culturally diverse language coverage, organizing resources by linguistic phenomena, domains, and modalities including speech and multimodal tasks.

arXiv cs.CL·Apr 20

58

Illustration for: Knowing When to Quit: A Principled Framework for Dynamic Abstention in LLM Reasoning

Knowing When to Quit: A Principled Framework for Dynamic Abstention in LLM Reasoning

Researchers formalize dynamic mid-generation abstention for LLMs using reinforcement learning, enabling models to halt unpromising reasoning chains early and reduce wasted compute. The framework models abstention as an explicit action with tunable trade-offs between computational cost and output quality.

arXiv cs.CL·Apr 20

58

Illustration for: CEO and CFO suddenly depart AI nuclear power upstart Fermi

Business & Funding Hardware & Infra

CEO and CFO suddenly depart AI nuclear power upstart Fermi

Fermi, the nuclear power startup co-founded by former Energy Secretary Rick Perry and positioned to power AI datacenters, lost its CEO and CFO amid operational challenges at its Texas campus. The dual departure signals trouble for a company betting on nuclear as the infrastructure backbone for AI scaling.

TechCrunch — AI·Apr 20

65

Illustration for: StepPO: Step-Aligned Policy Optimization for Agentic Reinforcement Learning

Research Models & Releases

StepPO: Step-Aligned Policy Optimization for Agentic Reinforcement Learning

Researchers propose StepPO, a reinforcement learning method designed to train LLMs for multi-turn agentic tasks like tool use and decision-making. The approach addresses challenges unique to agent training: sparse rewards, long contexts, and variable-length interactions that differ from single-turn alignment methods like RLHF.

arXiv cs.CL·Apr 20

62

Illustration for: AlphaContext: An Evolutionary Tree-based Psychometric Context Generator for Creativity Assessment

AlphaContext: An Evolutionary Tree-based Psychometric Context Generator for Creativity Assessment

Researchers introduce AlphaContext, a system that uses evolutionary algorithms to generate psychometric test scenarios for measuring creative thinking in LLM-era workflows. The tool aims to overcome limitations in existing AI-generated assessment contexts by improving narrative coherence and stylistic variety.

arXiv cs.CL·Apr 20

42

Illustration for: River-LLM: Large Language Model Seamless Exit Based on KV Share

Research Tools & Code

River-LLM: Large Language Model Seamless Exit Based on KV Share

River-LLM tackles a fundamental bottleneck in early-exit inference: when decoder-only models skip layers, downstream tokens lose access to cached key-value states, killing practical speedups. This training-free framework restores efficiency by sharing KV cache across exited layers, bridging the gap between theoretical and real-world latency gains.

arXiv cs.CL·Apr 20

58

Illustration for: Understanding the Prompt Sensitivity

Understanding the Prompt Sensitivity

Researchers used Taylor expansion to model LLM behavior as multivariate functions, revealing that large language models disperse rather than cluster similar inputs, leading to high variance in outputs for meaning-preserving prompts. The finding explains why LLMs exhibit prompt sensitivity and has implications for reliability in production systems.

arXiv cs.CL·Apr 20

58

Illustration for: IceBreaker for Conversational Agents: Breaking the First-Message Barrier with Personalized Starters

IceBreaker for Conversational Agents: Breaking the First-Message Barrier with Personalized Starters

Researchers propose Conversation Starter Generation, a technique to help conversational agents like ChatGPT and Doubao proactively initiate personalized dialogue with users who lack explicit queries. The work addresses a real deployment gap: how to bootstrap engagement when users have vague needs but no clear starting point.

arXiv cs.CL·Apr 20

52

Illustration for: ArbGraph: Conflict-Aware Evidence Arbitration for Reliable Long-Form Retrieval-Augmented Generation

Research Tools & Code

ArbGraph: Conflict-Aware Evidence Arbitration for Reliable Long-Form Retrieval-Augmented Generation

Researchers introduce ArbGraph, a framework that resolves factual conflicts in retrieval-augmented generation before text generation begins. The system decomposes retrieved documents into atomic claims, maps contradiction and support relations, and uses credibility propagation to arbitrate conflicts, addressing a core reliability problem in long-form RAG pipelines.

arXiv cs.CL·Apr 20

58

Illustration for: Omni-Embed-Audio: Leveraging Multimodal LLMs for Robust Audio-Text Retrieval

Research Models & Releases

Omni-Embed-Audio: Leveraging Multimodal LLMs for Robust Audio-Text Retrieval

Researchers propose Omni-Embed-Audio, a retrieval system that uses multimodal LLMs to improve audio-text search beyond caption-based queries. The work introduces User-Intent Queries spanning questions, commands, tags, and paraphrases to stress-test real-world robustness, plus new metrics for evaluating hard negative cases.

arXiv cs.CL·Apr 20

52

Illustration for: ComPASS: Towards Personalized Agentic Social Support via Tool-Augmented Companionship

ComPASS: Towards Personalized Agentic Social Support via Tool-Augmented Companionship

Researchers introduce ComPASS, a framework that equips conversational agents with external tools to deliver personalized emotional support across multiple modalities. The work grounds agentic companionship in psychological social support theory and includes ComPASS-Bench, a new benchmark for evaluating personalized support interactions.

arXiv cs.CL·Apr 20

52

Illustration for: PRISMA: Preference-Reinforced Self-Training Approach for Interpretable Emotionally Intelligent Negotiation Dialogues

PRISMA: Preference-Reinforced Self-Training Approach for Interpretable Emotionally Intelligent Negotiation Dialogues

Researchers introduce PRISMA, a dialogue system that combines emotion recognition with negotiation strategy to generate interpretable responses in job interviews and resource allocation scenarios. The work addresses a gap in emotionally aware AI by prioritizing explainability alongside emotional intelligence.

arXiv cs.CL·Apr 20

52

Illustration for: HiGMem: A Hierarchical and LLM-Guided Memory System for Long-Term Conversational Agents

Research Tools & Code

HiGMem: A Hierarchical and LLM-Guided Memory System for Long-Term Conversational Agents

Researchers propose HiGMem, a two-level memory architecture that uses LLM-guided event summaries to retrieve relevant conversation history more precisely than vector similarity alone. The system addresses a core challenge in long-context agents: bloated retrieval sets that waste context budget and degrade answer quality.

arXiv cs.CL·Apr 20

58

Illustration for: Multilingual Training and Evaluation Resources for Vision-Language Models

Research Tools & Code

Multilingual Training and Evaluation Resources for Vision-Language Models

Researchers released Multi-PixMo, a multilingual training and evaluation suite for vision-language models covering five European languages. The resource combines synthetic generation and manual annotation to address the scarcity of non-English VLM datasets and benchmarks, filling a gap in cross-lingual multimodal AI development.

arXiv cs.CL·Apr 20

54

Illustration for: Adobe fights AI disruption of its own business model with new enterprise agent platform

Products & Apps Business & Funding

Adobe fights AI disruption of its own business model with new enterprise agent platform

Adobe is launching an enterprise agent platform to counter AI-native startups threatening its core business, while simultaneously searching for a new CEO. The move signals how established software vendors are scrambling to defend market share against AI disruption.

The Decoder·Apr 20

73

Illustration for: Physical AI Edges Closer to Real-World Deployments

Business & Funding Research

Physical AI Edges Closer to Real-World Deployments

Capgemini's latest research shows enterprises are shifting physical AI systems from pilot programs into live operations at scale. The move signals that robotics and embodied AI are maturing beyond controlled lab settings into production environments where they face real operational constraints.

AI Business·Apr 20

61

Illustration for: Chinese Vendor Claims First Large-Scale Embodied AI Deployment

Products & Apps Business & Funding

Chinese Vendor Claims First Large-Scale Embodied AI Deployment

Agibot claims to have moved embodied AI from experimental pilots into real industrial deployment at scale, signaling a potential inflection point for robotics in manufacturing and logistics workflows.

AI Business·Apr 20

61

Illustration for: Google plans nearly two million new AI chips as it turns to Marvell for custom designs

Hardware & Infra Business & Funding

Google plans nearly two million new AI chips as it turns to Marvell for custom designs

Google is partnering with Marvell Technology to design two custom AI chips for its data centers, planning to deploy nearly two million units. The move signals Google's shift toward vertical integration of silicon to reduce reliance on third-party vendors and control costs at scale.

The Decoder·Apr 20

85

Illustration for: Salesforce bets on "Agent Albert" to prove AI won't kill enterprise software

Products & Apps Business & Funding

Salesforce bets on "Agent Albert" to prove AI won't kill enterprise software

Salesforce is launching Agent Albert, an AI product designed to counter Wall Street concerns that generative AI could render traditional enterprise software irrelevant. CEO Marc Benioff is also introducing a proprietary metric to demonstrate the technology's business value.

The Decoder·Apr 20

73

Illustration for: UK Launches $675 Million Fund for AI Startups

Business & Funding Policy & Regulation

UK Launches $675 Million Fund for AI Startups

The UK government committed $675 million to back AI startups through a new capital program, joining a wave of state-backed funding initiatives globally. The move signals sovereign competition for AI talent and infrastructure outside the US venture ecosystem.

AI Business·Apr 20

61

Illustration for: Import AI 454: Automating alignment research; safety study of a Chinese model; HiFloat4

Research Models & Releases

Import AI 454: Automating alignment research; safety study of a Chinese model; HiFloat4

Import AI 454 covers three substantive topics: automating alignment research to scale safety work, a safety evaluation of a Chinese LLM, and HiFloat4, a new numerical format for model training. The lead question on market pricing of AGI signals broader economic implications.

Import AI (Jack Clark)·Apr 20

77

Illustration for: US, California Use Purchasing Power to Set AI Rules

Policy & Regulation

US, California Use Purchasing Power to Set AI Rules

US federal and state governments are leveraging procurement rules to enforce AI governance where legislation hasn't materialized, with California and federal agencies using purchasing power as a de facto regulatory lever on AI vendors.

AI Business·Apr 20

61

Illustration for: Tech CEOs Think AI Will Let Them Be Everywhere at Once

Opinion & Analysis

Tech CEOs Think AI Will Let Them Be Everywhere at Once

Zuckerberg and Dorsey are exploring AI systems that would amplify executive control and presence across their organizations, raising questions about how AI-driven management tools reshape corporate power dynamics.

WIRED — AI·Apr 20

58

Illustration for: Towards Real-Time ECG and EMG Modeling on $μ$ NPUs

Research Hardware & Infra

Towards Real-Time ECG and EMG Modeling on $μ$ NPUs

PhysioLite enables real-time ECG and EMG analysis on microcontroller-scale neural processing units by replacing Transformer attention with learnable wavelet filters and hardware-aware layer design, solving a gap where physiological signal processing was previously infeasible on edge devices.

arXiv cs.LG·Apr 20

58

Older stories →