AI Research Paper Archive & Publications
Browse through 346 peer-reviewed AI publications, arxiv preprints, technical abstracts, and key takeaway summaries covering LLM architectures, vision models, alignment, and multi-agent systems.
By Mattia Greiche, Reihaneh Rabbany, Maximilian Puelma Touzel
Controlled context enrichment (neutral, social, descriptive, normative) and selective robustness (invariance to paraphrase + directional sensitivity) reveal distinct behavioral shifts in LLM social agents across two models.
By Nicolas Tricard, Linzheng Wang, Jason Chen et al.
Benchmark of 10 learned vs 5 classical flow-reconstruction methods shows the best approach shifts per regime; classical estimators win in dense/low-order sensing while generative models win as sensor density drops.
By Jisoo Kim, Chaehyun Kim, Junsu Kim et al.
DreamX repurposes a pretrained video diffusion model to extract compact, embodiment-invariant latent actions from video, enabling zero-shot action transfer and efficient policy learning across humans and robots.
By Hao Fei
Proposes cognitive symbionts—lifelong, embedded AI companions that augment human intelligence by closing the cognitive gap between potential and working memory.
By Jason Jingzhou Liu, Tony Tao, Deepak Pathak et al.
Pixel-UMI learns whole-body robot policies from sensorized human demos via collision-aware environment retargeting, viewpoint-error masking to bridge embodiment gaps, and DAgger on-policy human corrections.
By Tony Tao, Jason Jingzhou Liu, Ruslan Salakhutdinov et al.
Pixel is a full-stack co-designed robot system enabling general-purpose mobile manipulation via low-cost hardware, scalable data collection, unified control, and hierarchical policy learning.
By Byungwoo Jeon, Yoonwoo Jeong, Hyunseok Lee et al.
LARC enables unified multimodal models to reason in continuous latent space rather than discrete tokens, using curriculum SFT and information-gain RL to improve text-to-image generation and achieve test-time scaling.
By Bobo Li, Jiajia Song, Hao Fei et al.
Flide decouples logical tree structure from physical execution using ForkAttention to exactly navigate branches in flattened token streams, achieving up to 9.5× faster decision inference than vLLM and SGLang.
By Nicolas Dufour, Elouan Gardès, Amil Dravid et al.
Neural Thicket gains come from output diversity, not weight geometry: freezing the model and adding random logit biases matches full-weight perturbation on reasoning tasks.
By Quentin Dhôte, Romaric Audigier
A multi-stream video luggage re-identification pipeline that merges multiple shots and views, trained on diverse real-world environments, improving robustness to occlusions and clutter.
By Jiongfan Yang, Weichen Fan, Haiwen Diao et al.
SpecSlot decouples output detail from global attention length by expanding each spatial anchor into multiple DCT-band slots, enabling faster convergence and better FID with fewer FLOPs.
By Aron Brenner, Saurabh Amin
Centering KL-DRO ambiguity set at a learned density model instead of empirical distribution enables extrapolation to costly unobserved scenarios, with adaptive importance-weighted scenario generation for tractable solutions.
Frontier Artificial Intelligence Research Disciplines
Our research archive systematically categorizes preprints and conference proceedings across core theoretical and applied machine learning domains:
Language Models & Scaling
Investigating transformer scaling laws, sparse Mixture-of-Experts (MoE), test-time compute scaling, reasoning token generation, and deep reasoning architectures.
Vision, Audio & Video
Advancements in unified multimodal tokenizers, autoregressive video synthesis, latent diffusion models, and real-time spatial scene understanding.
Alignment & Interpretability
Techniques in Direct Preference Optimization (DPO), Reinforcement Learning from AI Feedback (RLAIF), sparse autoencoders, and mechanistic circuit auditing.
Autonomous Agents & Tool Use
Multi-agent coordination protocols, external environment simulation, self-correcting code synthesis, and long-horizon planning frameworks.
Inference Optimization
Hardware-aware attention mechanisms, FlashAttention kernels, sub-4-bit quantization (AWQ/GPTQ), speculative decoding, and KV cache eviction policies.
Benchmark Integrity & Contamination
Methodologies addressing test-set leakage, dynamic synthetic benchmark generation, human-in-the-loop validation, and adversarial capability probes.
Frequently Asked Questions on AI Research
How frequently is the research paper archive updated?▼
Our indexing pipeline scans daily arXiv submissions in cs.CL, cs.AI, cs.CV, and cs.LG alongside major conference proceedings (NeurIPS, ICML, ICLR, ACL). Publications are updated continuously with fresh abstracts and takeaways.
Where can I access the original full-text PDF of a paper?▼
Every individual paper synopsis page on llmdb.app contains direct, verified links to the original arXiv abstract, PDF download mirror, and official GitHub source code repositories when published by the authors.