Carlos's Debrief

April 23, 2026 09:00
44ArXiv Papers
35Web Findings
79Total Sources
Topics searched: Artificial Intelligence Machine Learning Large Language Models Security & Cybersecurity Cryptography Zero Knowledge Quantum Computing Crypto & Blockchain AI Agents & Reasoning AI Safety & Alignment
April 23, 2026 09:00 — Curated by Hermes Research Scout

⚡ Quick Summary

📰 News Headlines

📄 ArXiv Papers

🧠 LLMs 7

Lukas Hof, Pavel Kroupa, Gerhard Hensler, Jan Pflamm-Altenburg
Semi-analytical evolution models of galaxies with metallicity and SFR-dependent IGIMF on star-formation self-regulation show IGIMF theory produces reasonable gas fractions and depletion timescales.
astro-ph.GA
Ruohan Liu, Shukang Yin, Tao Wang, Dong Zhang, Weiji Zhuang, Shuhuai Ren, Ran He, Caifeng Shan, Chaoyou Fu
Paralinguistic cues are essential for natural human-computer interaction, yet their evaluation in Large Audio-Language Models (LALMs) remains limited by coarse feature coverage and the inherent subjectivity of assessment. To address…
cs.CL cs.AI cs.SD
Clifford Taubes, Yingying Wu
Novel local singularity models for Z/2 harmonic 1-forms, self-dual 2-forms, and spinors in dimension 4 are described as homogeneous versions on R^4 whose singular sets are cones on the 1-skeletal…
math.DG math.AP
Zhaofeng Wu, Shiqi Wang, Boya Peng, Anuj Goyal, Melanie Kambadur, Sebastian Ruder, Yoon Kim, Chloe Bi
Parallel-SFT enables zero-shot cross-programming-language transfer for code RL, allowing capabilities acquired in higher-resource PLs to transfer to lower-resource PLs even without parallel training data.
cs.CL
Hanqi Li, Lu Chen, Kai Yu
As LLMs are increasingly integrated into agentic systems, they must adhere to dynamically defined machine-interpretable interfaces. RoboGrid evaluates LLMs as in-context interpreters through controlled stress-tests of recursion depth, expression complexity…
cs.AI
Pavel Salovskii, Iuliia Gorshkova
A hybrid architecture extends LLMs with an external ontological memory layer using RDF/OWL representations for persistent, verifiable, and semantically grounded reasoning, replacing vector-based retrieval with structured knowledge graph construction.
cs.AI
Mariano Barone, Francesco Di Serio, Roberto Moio, Marco Postiglione, Giuseppe Riccio, Antonio Romano, Vincenzo Moscato
A multidimensional evaluation of LLMs across structured medical explanations and real-world physician-patient interactions reveals baseline models amplify affective polarity relative to physicians, raising concerns about clinical deployment of AI assistants.
cs.CL cs.AI

⚙️ Machine Learning 13

Hyeonwoo Kim, Jeonghwan Kim, Kyungwon Cho, Hanbyul Joo
DeVI leverages text-conditioned synthetic videos for physically plausible dexterous robot control using hybrid 3D/2D tracking rewards, enabling zero-shot generalization across diverse objects and interaction types.
cs.CV
Yupeng Zheng, Xiang Li, Songen Gu, Yuhang Zheng, Shuai Tian, Weize Li, Linbo Wang, Senyu Fei, Pengfei Li, Yinfeng Gao, Zebin Xing, Yilun Chen, Qichao Zhang, Haoran Li, Wenchao Ding
PokeVLA is a lightweight foundation model for embodied manipulation that infuses vision-language understanding into action learning via a two-stage training paradigm, achieving strong performance on pocket-sized hardware.
cs.RO
Sina Gholami, Abdulmoneam Ali, Tania Haghighi, Ahmed Arafa, Minhaj Nur Alam
FedSIR is a multi-stage framework for robust federated learning under noisy labels that leverages spectral structure of client feature representations to identify mislabeled clients and relabel them, improving performance without…
cs.LG cs.AI cs.CV cs.DC eess.SP
Ana Sanchez-Fernandez, Thomas Pinetz, Werner Zellinger, Günter Klambauer
Batch effects systematically undermine deep learning on new experimental batches in biomedical imaging. Control-Stabilized Adaptive (CSA) uses in-context control samples to close the domain gap across heterogeneous lab conditions.
cs.LG q-bio.QM
Thorsten Hoeser, Felix Bachofer, Claudia Kuenzer
A global Sentinel-1 SAR time series dataset enables high-temporal-resolution monitoring of offshore wind infrastructure deployment and operation at global scale, filling a critical gap in temporally dense, semantically fine-grained offshore…
cs.CV cs.LG
Yiming Bian, Joshua M. Akey
CQS Divide, derived from cyclic quorum sets theory, decomposes attention into composable operations that remove the assumption that full QKV tensors fit in device memory, enabling long-context LLMs without OOM…
cs.LG cs.DC
Deqing Fu, Tianyi Zhou, Mikhail Belkin, Vatsal Sharan, Robin Jia
Language models trained on natural text learn periodic number features with dominant periods at T=2, 5, 10. This paper identifies a two-tiered hierarchy: only some models learn geometrically separable features…
cs.CL cs.AI cs.LG
Shelly Golan, Michael Finkelson, Ariel Bereslavsky, Yotam Nitzan, Or Patashnik
ParetoSlider enables inference-time control over inherently conflicting goals during diffusion model post-training, replacing fixed scalarization with a Pareto-based approach that maintains multiple trade-off points.
cs.LG cs.CV
Qiguang Chen, Chengyu Luan, Jiajun Wu, Qiming Yu, Yi Yang, Yizhuo Li, Jingqi Tong, Xiachong Feng, Libo Qin, Wanxiang Che
OMIBench evaluates Olympiad-level reasoning when required evidence is distributed over multiple images, spanning biology, chemistry, mathematics, and physics. It reveals that current LVLMs fail to exploit cross-image contextual information that…
cs.CV cs.AI cs.CL
Dimitrije Antić, Alvaro Budria, George Paschalidis, Sai Kumar Dwivedi, Dimitrios Tzionas
InterFields encodes dense, continuous proximity across body and object surfaces to reconstruct 3D Human-Object Interaction from a single RGB image, going beyond sparse binary contact cues to model subtle physical…
cs.CV cs.LG
Pranava Madhyastha, Dagmar Adamcova
Human-like working memory constraints integrated into Transformers via fixed-width windows and temporal decay attention variants enable modified GPT-2 models trained on developmentally plausible datasets to align with human reading time…
cs.CL cs.AI cs.LG
Zesheng Liu, Maryam Rahnemoonfar
Graph-based models for ice stratigraphy extended to handle incomplete radar layer traces and entirely missing layers using physics-conditioned synthesis, enabling prediction of deeper-layer thickness from partially observed shallow layers.
cs.LG
Chao Wang, Luca Nepote, Giulio Franzese, Pietro Michiardi
A general framework for estimating KL divergence between probability distributions in function space, with applications to trajectory inference in single-cell genomics where path-space laws are non-identifiable from finitely many marginals.
cs.LG

🔒 Security & Crypto 10

Mikko Lempinen, Joni Kemppainen, Niklas Raesalmi
AVISE (AI Vulnerability Identification and Security Evaluation) is a modular open-source framework for identifying vulnerabilities in and evaluating the security of AI systems and models, addressing the growing risks as…
cs.CR cs.AI cs.CL
Alexander Berladskyy, Andreas Aßmuth
Phishing attacks exceeded 890,000 in Q3 2025. This paper examines FIDO2/passkeys claiming phishing-resistant authentication, identifying new attack vectors that challenge these claims as passkey adoption scales.
cs.CR
Travis LaCroix
The value alignment problem is reconceptualized as a structural governance question using the principal-agent framework from economics — not whether an AI is aligned in the abstract, but whether it…
cs.CY cs.AI cs.MA
Hanzhi Liu, Chaofan Shou, Xiaonan Liu, Hongbo Wen, Yanju Chen, Ryan Jingyang Fang, Yu Feng
LLM agents find vulnerabilities that human auditors miss, but the harness — how roles, tools, and retries are wired — dramatically affects success rates even with a fixed language model.
cs.CR
Ray Iskander, Khaled Kirah
Machine-checked proofs establish that per-stage arithmetic masking guarantees pipeline-level security for post-quantum cryptographic accelerators (ML-KEM, ML-DSA), resolving a long-standing composition gap.
cs.CR
Shahid Alam, Amina Jameel, Zahida Parveen, Ehab Alnfrawy, Adeela Ashraf, Raza Uddin, Jamal Aqib
DAIRE is a lightweight ML framework for real-time detection and classification of Controller Area Network (CAN) attacks in the Internet of Vehicles, addressing critical automotive communication protocol vulnerabilities.
cs.CR cs.AI
Lena Sinterhauf, Andreas Aßmuth, Roland Kaltefleiter
Analysis of 245,456 CVEs with CVSS scores >= 9.0 reveals patterns in identification and resolution timelines, providing empirical guidance for prioritizing remediation of the most critical vulnerabilities.
cs.CR
Abhijit Talluri
Auto-ART provides structured synthesis of nine peer-reviewed corpus sources (2020-2026) plus an open-source automated adversarial robustness testing framework, addressing fragmented evaluation protocols in trustworthy ML.
cs.CR cs.LG
Shahriar Rahman Khan, Tariqul Islam, Raiful Hasan
This SoK systematizes 48 peer-reviewed studies on perception-layer attacks against autonomous vehicles, developing a unified taxonomy of 20 attack vectors organized by sensor type, attack stage, medium, and threat model.
cs.CR
Jianxin Gao, Ruohan Lei, Wanli Peng
Multimodal LLMs enable text steganography that addresses both white-box exposure risk (via dynamic codebook) and black-box inflexibility (via per-message extracting prompts), improving both security and practicality.
cs.CR cs.AI

⚛️ Quantum 8

Jie Gu, Jue Hou, Yunfeng Jiang
Harmonic analysis via Maass waveforms and Eisenstein series provides a numerically stable method to compute T-Tbar-deformed torus partition functions at finite deformation parameter λ.
hep-th math-ph
Koki Okada, Kenta Kasai
A quantum LDPC code family is constructed from a length-512 Calderbank-Shor-Steane base matrix pair that is (3,8)-regular with girth 8. CPM lifts produce codes with parameters [[512, 174, 8]].
quant-ph
Justin Tim-Lok Chau, Jiarui Zhao, Nicolas Laflorencie, Zi Yang Meng
Quantum Monte Carlo with split-spin representation resolves the continuous quantum phase transition between gapped Haldane phase and gapless long-range interacting spin-1 chain, revealing unconventional critical behavior.
cond-mat.str-el quant-ph
Sergei Igonin
Semi-discrete matrix Lax representations (Lax pairs) are studied in integrable differential-difference equations. Gauge transformations of MLRs are analyzed and their invariants are characterized.
nlin.SI
Nitay Hurvitz, Alon Kochol, Victor Fleurov, Eran Sela
The arrow of time (AoT) — log ratio of forward/backward trajectory probabilities — is introduced as a thermodynamic indicator of Measurement-Induced Phase Transitions (MIPTs), complementing entanglement-based measures.
cond-mat.stat-mech quant-ph
Ali Rayat, Yaohang Li, Gia-Wei Chern
A gauge-equivariant graph neural network embeds non-Abelian symmetry directly into message passing via matrix-valued, gauge-covariant features, extending equivariant learning to intrinsically nonlocal observables.
cond-mat.str-el cs.LG hep-lat
Cristian D. Suarez-Jimenez, Alejandro P. Riascos, Denis Boyer
Exact analytical expressions are derived for mean first-encounter time of multiple random walkers on networks under simultaneous resetting protocol, expressed in terms of eigenvalues and eigenvectors of the transition matrix.
cond-mat.stat-mech
Víctor Zapatero, Marcos Curty
Device-independent quantum cryptography is analyzed under partial input leakage modeled as a noisy channel. Results quantify certifiable local randomness and secret key rate as a function of input leakage magnitude.
quant-ph

🛡️ AI Safety 1

Carl Angelo Angcana, Jamlech Iram Gojo Cruz
The Cat Distribution System (CDS) folklore — that cats 'find' their owners — is used as a lens to redesign adoption platforms in the Philippines, where 13.11 million stray cats…
cs.HC

📦 Other 5

Bernhard Haeupler, Slobodan Mitrović, Srikkanth Ramachandran, Wen-Horng Sheu, Robert Tarjan
A wide class of local search algorithms is extended to the fully dynamic setting with an adaptive adversary, achieving amortized O-tilde(1) local-search steps per update using witness-tree and entropy compression.
cs.DS
Bob Wilson
For minimax problems with bilinear objectives, the proximal operator reduces to a generalized projection onto the confidence region, enabling efficient solution via ADMM without the usual computational hardness.
math.OC stat.ME
James L. Banal
Synthetic DNA approaches 227.5 exabytes per gram with millennial stability. A new coding scheme retains the sequencer's per-position posterior distribution for error correction, approaching the information-theoretic capacity limit.
cs.IT
David Muñoz-Lahoz
Archimedean lattice-ordered algebras with identity that admit a polynomial growth continuous function calculus are characterized, generalizing functional calculus results to the lattice-ordered algebra setting.
math.FA
Erik Pelttari, Selda Kücükçifççi, E. Şule Yazıcı
Heffter arrays are combinatorial structures for constructing orthogonal cyclic cycle decompositions. This paper constructs globally simple Heffter arrays H(n;k) when k ≡ 1 (mod 4), extending existing results.
math.CO

🌐 Web Findings

🦞 Lobste.Rs 3

A technical analysis reversing Google's SynthID watermarking system for AI-generated images, exploring how the embedding mechanism works and potential limitations of synthetic watermarking approaches.
Lobste.rs
A deep-dive into how North Korea's Lazarus Group leverages AI to scale and industrialize attacks targeting open-source software developers, including AI-assisted social engineering, malicious code…
Lobste.rs
A developer decrypted the mechanism by which Cloudflare intercepts and reads client-side React state in the ChatGPT interface before user keystrokes reach OpenAI, raising significant…
Lobste.rs

🤗 Huggingface Papers 24

DeVI leverages text-conditioned synthetic videos for physically plausible dexterous robot control using hybrid 3D/2D tracking rewards, enabling zero-shot generalization across diverse objects.
HuggingFace
TACO automatically discovers and refines compression rules from interaction trajectories, reducing token overhead by ~10% while improving performance 1-4% on TerminalBench.
HuggingFace
Tadabur provides 1,400+ hours of Quran recitation audio from 600+ reciters for Quranic speech research, significantly expanding both scale and diversity of available Quranic data.
HuggingFace
Across 25,000+ LLM agent runs, evidence is ignored in 68% of traces and refutation-driven belief revision occurs in only 26% — current agents execute scientific…
HuggingFace
In settings where full latent identifiability is unattainable, intersections, complements, and symmetric differences of latent variables are shown to be identifiable up to appropriate indeterminacies.
HuggingFace
SAVOIR uses Shapley values for fair credit distribution in social intelligence tasks; a 7B model matches or exceeds GPT-4o and Claude-3.5-Sonnet on the SOTOPIA benchmark.
HuggingFace
ReImagine decouples appearance modeling from temporal consistency by learning a high-quality image prior then using it for video synthesis, producing temporally consistent videos under diverse…
HuggingFace
ToolsRL uses direct tool supervision via simple visual tools with a curriculum that masters tool-calling before using tools for complex visual reasoning tasks.
HuggingFace
Recursive Tournament Voting (RTV) over compact rollout trajectory summaries boosts Claude-4.5-Opus from 70.9% to 77.6% on SWE-Bench Verified for agentic coding.
HuggingFace
NPO learns from a policy's own future checkpoint, which is both stronger and closer than any external source. On Qwen3-VL-8B-Instruct with GRPO, NPO improves 57.88…
HuggingFace
The Proxy Compression Hypothesis (PCH) unifies reward hacking in RLHF/RLAIF/RLVR regimes, explaining how objective compression and optimization amplification cause local shortcut learning to generalize into…
HuggingFace
GSI-Bench quantifies generative spatial intelligence in multimodal LLMs via spatially grounded image editing, providing first clear evidence that generative training strengthens spatial reasoning.
HuggingFace
Transformers, RNNs, and word embeddings independently learn periodic number features (T=2,5,10). Fourier domain sparsity is necessary but not sufficient for geometric separability.
HuggingFace
WavAlign makes RL practical for end-to-end spoken dialogue by constraining preference updates to the semantic channel and improving acoustic behavior via explicit anchoring.
HuggingFace
CluE groups heterogeneous memory extraction tasks by scenario and synthesizes cross-cluster insights, achieving +9.04% relative gain over prior self-evolving frameworks.
HuggingFace
LLaDA2.0-Uni is a unified discrete diffusion LLM combining a MoE dLLM backbone with a diffusion decoder, matching specialized VLMs while delivering strong image generation.
HuggingFace
DR-Venus is a frontier 4B deep research agent built entirely on ~10K open-data, outperforming prior agentic models under 9B parameters and narrowing the gap to…
HuggingFace
MMCORE transfers VLM semantic understanding into visual generation by predicting visual embeddings via learnable query tokens, eliminating deep fusion between autoregressive and diffusion models.
HuggingFace
CreativeGame uses multi-agent iterative game generation with programmatic reward, lineage-scoped memory, runtime validation, and mechanic-guided planning for interpretable version-to-version evolution.
HuggingFace
Cortex 2.0 shifts from reactive VLA control to plan-and-act via candidate future trajectory generation in visual latent space, consistently outperforming state-of-the-art VLA baselines in complex…
HuggingFace
SWE-chat captures 6,000 real coding agent sessions with 63,000 user prompts. Only 44% of agent code survives into user commits; AI-authored code introduces more security…
HuggingFace
SkillLearnBench benchmarks continual skill learning for LLM agents across 20 verified tasks, finding all methods improve over baselines but no method leads across all tasks…
HuggingFace
Stargazer evaluates AI agents on dynamic radial-velocity model-fitting tasks; frontier agents achieve good statistical fits but frequently fail to recover correct physical system parameters.
HuggingFace
SmartPhotoCrafter tightly couples an Image Critic module (identifying deficiencies) with a Photographic Artist module (realizing targeted edits), eliminating the need for explicit human aesthetic instructions.
HuggingFace

📚 Google Scholar 8

A framework for identifying breakthrough technologies using machine learning-based link prediction to detect innovations with potential to trigger technological breakthroughs.
Google Scholar
Machine learning approaches are applied to methane dry reforming catalyst design to address rising atmospheric CO2 and methane levels through advanced carbon capture techniques.
Google Scholar
A review of LLMs in computational social science examining prospects, current state, and challenges around data bias, privacy, and model integration.
Google Scholar
A comprehensive review covering adversarial AI, automated threat intelligence, and AI-driven security orchestration as the key areas defining AI's expanding role in cybersecurity.
Google Scholar
With quantum computers threatening traditional cryptography, post-quantum solutions are essential for securing IoT devices and user privacy against future quantum attacks.
Google Scholar
A comprehensive review of quantum machine learning spanning quantum-enhanced classical ML, native quantum algorithms, and hybrid approaches for optimization, drug discovery, and quantum-secured communications.
Google Scholar
Quantum algorithms applied to enhance efficiency, accuracy, and scalability across the automotive value chain, exploring recent advances in quantum computing.
Google Scholar
A framework integrating AI for rapid cyber threat analysis, blockchain for immutable compliance logging, and smart contracts for automated security policy enforcement.
Google Scholar

🔗 All Sources

  1. [1] SpeechParaling-Bench: A Comprehensive Benchmark for Paralinguistic-Aware Speech Generation
  2. [2] AVISE: Framework for Evaluating the Security of AI Systems
  3. [3] FedSIR: Spectral Client Identification and Relabeling for Federated Learning with Noisy Labels
  4. [4] Convergent Evolution: How Different Language Models Learn Similar Number Representations
  5. [5] Diagnosing CFG Interpretation in LLMs
  6. [6] OMIBench: Benchmarking Olympiad-Level Multi-Image Reasoning in Large Vision-Language Model
  7. [7] Relative Principals, Pluralistic Alignment, and the Structural Value Alignment Problem
  8. [8] Automatic Ontology Construction Using LLMs as an External Layer of Memory, Verification, and Planning
  9. [9] Can "AI" Be a Doctor? A Study of Empathy, Readability, and Alignment in Clinical LLMs
  10. [10] Working Memory Constraints Scaffold Learning in Transformers under Data Scarcity
  11. [11] Closing the Domain Gap in Biomedical Imaging by In-Context Control Samples
  12. [12] Global Offshore Wind Infrastructure: Deployment and Operational Dynamics from Dense Sentinel-1 Time Series
  13. [13] Stream-CQSA: Avoiding Out-of-Memory in Attention Computation via Flexible Workload Scheduling
  14. [14] ParetoSlider: Diffusion Models Post-Training for Continuous Reward Control
  15. [15] LEXIS: LatEnt ProXimal Interaction Signatures for 3D HOI from an Image
  16. [16] Gauge-Equivariant Graph Neural Networks for Lattice Gauge Theories
  17. [17] Physics-Conditioned Synthesis of Internal Ice-Layer Thickness for Incomplete Layer Traces
  18. [18] Self-regulated galaxy evolution within a self-consistently varying galaxy-wide IMF
  19. [19] DeVI: Physics-based Dexterous Human-Object Interaction via Synthetic Video Imitation
  20. [20] Homogeneous Z/2-Harmonic Forms and Spinors on R^4 from Regular 4-Polytopes
  21. [21] High-Girth Regular Quantum LDPC Codes from Affine-Coset Structures
  22. [22] Parallel-SFT: Improving Zero-Shot Cross-Programming-Language Transfer for Code RL
  23. [23] PokeVLA: Empowering Pocket-Sized Vision-Language-Action Model with World Knowledge Guidance
  24. [24] An Analysis of Attack Vectors Against FIDO2 Authentication
  25. [25] Synthesizing Multi-Agent Harnesses for Vulnerability Discovery
  26. [26] Fresh Masking Makes NTT Pipelines Composable: Machine-Checked Proofs for PQC Hardware
  27. [27] DAIRE: A lightweight AI model for real-time CAN attack detection in the Internet of Vehicles
  28. [28] CVEs With a CVSS Score Greater Than or Equal to 9
  29. [29] Auto-ART: Structured Literature Synthesis and Automated Adversarial Robustness Testing
  30. [30] SoK: The Next Frontier in AV Security: Systematizing Perception Attacks and Multi-Sensor Fusion Threats
  31. [31] Device-independent quantum cryptography with input leakage
  32. [32] Lattice-ordered algebras admitting a polynomial growth continuous function calculus
  33. [33] Text Steganography with Dynamic Codebook and Multimodal Large Language Model
  34. [34] Globally Simple Heffter Arrays H(n;k) with k ≡ 1 (mod 4)
  35. [35] The gauge action on semi-discrete Lax representations and its invariants
  36. [36] Beyond Hagedorn: A Harmonic Approach to T-Tbar-deformation
  37. [37] Dynamic Construction of the Lovász Local Lemma
  38. [38] Solving Minimax Problems with Bilinear Objectives with ADMM
  39. [39] Unconventional Quantum Criticality in Long-Range Spin-1 Chains
  40. [40] Arrow of Time as an indicator of Measurement-Induced Phase Transitions
  1. [41] DNA storage approaching the information-theoretic ceiling
  2. [42] Relative Entropy Estimation in Function Space: Theory and Applications to Trajectory Inference
  3. [43] Encounter times of random walkers with simultaneous resetting on networks
  4. [44] From Meme to Method: Rethinking Animal Adoption Platforms through the Cat Distribution System
  5. [45] Reversing SynthID
  6. [46] Inside Lazarus: How North Korea uses AI to industrialize attacks on developers
  7. [47] ChatGPT Won't Let You Type Until Cloudflare Reads Your React State
  8. [48] DeVI: Physics-based Dexterous Human-Object Interaction
  9. [49] TACO: Terminal Agent Compression via Self-Evolving Context Rules
  10. [50] Tadabur: A Large-Scale Quran Audio Dataset
  11. [51] AI scientists produce results without reasoning scientifically
  12. [52] Diverse Dictionary Learning
  13. [53] SAVOIR: Social RL via Shapley Value Attribution
  14. [54] ReImagine: Image-First Human Video Generation
  15. [55] Visual Reasoning through Tool-supervised RL
  16. [56] Scaling Test-Time Compute for Agentic Coding
  17. [57] Near-Future Policy Optimization (NPO)
  18. [58] Reward Hacking in the Era of Large Models
  19. [59] GSI-Bench: Generative Spatial Intelligence
  20. [60] Convergent Number Representations in Language Models
  21. [61] WavAlign: Adaptive Post-Training for Spoken Dialogue Models
  22. [62] Self-Evolving LLM Memory Extraction
  23. [63] LLaDA2.0-Uni: Unified Diffusion LLM for Multimodal Understanding & Generation
  24. [64] DR-Venus: Edge-Scale Deep Research with 10K Open Data
  25. [65] MMCORE: Multimodal Image Generation via VLM-Aligned Embeddings
  26. [66] CreativeGame: Mechanic-Aware Creative Game Generation
  27. [67] Cortex 2.0: World Models for Industrial Robot Manipulation
  28. [68] SWE-chat: Real Coding Agent Sessions (6K sessions, 63K prompts)
  29. [69] SkillLearnBench: Continual Skill Learning for Agentic LLMs
  30. [70] Stargazer: AI Agents for Astrophysical Model-Fitting
  31. [71] SmartPhotoCrafter: Reasoning-to-Generation Photographic Editing
  32. [72] Early identification of breakthrough technologies via ML
  33. [73] Catalyst breakthroughs in methane dry reforming via ML
  34. [74] Large language models in computational social science
  35. [75] AI and ML in cybersecurity: state-of-the-art and future paradigms
  36. [76] Post-quantum cryptography for IoT authentication and privacy
  37. [77] Quantum machine learning: integrating AI with quantum computing
  38. [78] Quantum computing for automotive applications
  39. [79] Automated cybersecurity compliance via AI, blockchain and smart contracts