Research

Research papers, whitepapers, case studies, etc. 1271 public listings.

• Countering Reward Over-optimization in LLM with Demonstration-Guided Reinforcement Learning (Cohere)
  While Reinforcement Learning (RL) has been proven essential for tuning large language models (LLMs), it can lead to reward over-optimizatio…
• Periodic agent-state based Q-learning for POMDPs (Cohere)
  The standard approach for Partially Observable Markov Decision Processes (POMDPs) is to convert them to a fully observed belief-state MDP.
• Déjà Vu: Multilingual LLM Evaluation through the Lens of Machine Translation Evaluation (Cohere)
  Generation capabilities and language coverage of multilingual large language models (mLLMs) are advancing rapidly.
• Procedural Knowledge in Pretraining Drives Reasoning in Large Language Models (Cohere)
  Nov 20, 2024 Procedural Knowledge in Pretraining Drives Reasoning in Large Language Models Authors Laura Ruis, Maximilian Mozes, Juhan Bae,…
• The Grand Illusion: The Myth of Software Portability and Implications for ML Progress (Cohere)
  Pushing the boundaries of machine learning often requires exploring different hardware and software combinations.
• Improving Reward Models with Synthetic Critiques (Cohere)
  Reward models (RMs) play a critical role in aligning language models through the process of reinforcement learning from human feedback.
• NeoBabel: A Multilingual Open Tower for Visual Generation (Cohere)
  Text-to-image generation advancements have been predominantly English-centric, creating barriers for non-English speakers and perpetuating…
• Fishing for Magikarp: Automatically Detecting Under-trained Tokens in Large Language Models (Cohere)
  The disconnect between tokenizer creation and model training in language models allows for specific inputs, such as the infamous Solid Gold…
• How to Improve the Robustness of Closed-Source Models on NLI (Cohere)
  Closed-source Large Language Models (LLMs) have become increasingly popular, with impressive performance across a wide range of natural lan…
• State of Cloud Native Development (Cncf)
  Cloud native is the foundation of modern software.
• Critical Learning Periods: Leveraging Early training Dynamics for Efficient Data Pruning (Cohere)
  Neural Machine Translation models are extremely data and compute-hungry.
• The Art of Asking: Multilingual Prompt Optimization for Synthetic Data (Cohere)
  Synthetic data has become a cornerstone for scaling large language models, yet its multilingual use remains bottlenecked by translation-bas…
• Robust Distillation for Worst-class Performance (Cohere)
  Knowledge distillation has proven to be an effective technique in improving the performance a student model using predictions from a teache…
• BAM! Just Like That: Simple and Efficient Parameter Upcycling for Mixture of Experts (Cohere)
  The Mixture of Experts (MoE) framework has become a popular architecture for large language models due to its superior performance over den…
• Adaptation Odyssey in LLMs: Why Does Additional Pretraining Sometimes Fail to Improve? (Cohere)
  In the last decade, the generalization and adaptation abilities of deep learning models were typically evaluated on fixed training and test…
• Elo Uncovered: Robustness and Best Practices in Language Model Evaluation (Cohere)
  In Natural Language Processing (NLP), the Elo rating system, originally designed for ranking players in dynamic games such as chess, is inc…
• No News is Good News: A Critique of the One Billion Word Benchmark (Cohere)
  Araújo, Jeffrey Hui, Nicholas Frosst Abstract The One Billion Word Benchmark is a dataset derived from the WMT 2011 News Crawl, commonly us…
• The Multilingual Alignment Prism: Aligning Global and Local Preferences to Reduce Harm (Cohere)
  A key concern with the concept of "alignment" is the implicit question of "alignment to what?".
• SnapKV: LLM Knows What You are Looking for Before Generation (Cohere)
  Apr 22, 2024 SnapKV: LLM Knows What You are Looking for Before Generation Authors Yuhong Li, Yingbing Huang, Bowen Yang, Bharat Venkitesh,…
• Policy Primer - Translating Safety (Cohere)
  Global AI safety efforts have gained traction and momentum, but a critical challenge remains: how to ensure safety across diverse languages…
• From One to Many: Expanding the Scope of Toxicity Mitigation in Language Models (Cohere)
  To date, toxicity mitigation in language models has almost entirely been focused on single-language settings.
• Repetition In Repetition Out: Towards Understanding Neural Text Degeneration from the Data (Cohere)
  There are a number of diverging hypotheses about the neural text degeneration problem, i.e., generating repetitive and dull loops, which ma…
• State of Cloud Native Development in India 2026 (Cncf)
  India is rapidly becoming one of the world’s most important cloud native markets. 2.25 million developers in India are now cloud native, wi…
• Policy Primer - Efficient AI (Cohere)
  The growing global demand for digital products and services is increasing energy needs and their associated climate impacts.
• Understanding and Mitigating Language Confusion in LLMs (Cohere)
  We investigate a surprising limitation of LLMs: their inability to consistently generate text in a user's desired language.
• FAIR-Ensemble: When Fairness Naturally Emerges From Deep Ensembling (Cohere)
  Ensembling independent deep neural networks (DNNs) is a simple and effective way to improve top-line metrics and to outperform larger singl…
• Metadata Archaeology: Unearthing Data Subsets by Leveraging Training Dynamics (Cohere)
  Modern machine learning research relies on relatively few carefully curated datasets.
• Contrastive Policy Gradient: Aligning LLMs on sequence-level scores in a supervised-friendly fashion (Cohere)
  Reinforcement Learning (RL) has been used to finetune Large Language Models (LLMs) using a reward model trained from preference data, to be…
• Exploring Low Rank Training of Deep Neural Networks (Cohere)
  Training deep neural networks in low rank, i.e.
• Investigating Continual Pretraining in Large Language Models: Insights and Implications (Cohere)
  This paper studies the evolving domain of Continual Learning (CL) in large language models (LLMs), with a focus on developing strategies fo…
• Interlocking Backpropagation: Improving depthwise model-parallelism (Cohere)
  The number of parameters in state of the art neural networks has drastically increased in recent years.
• To Code, or Not To Code? Exploring Impact of Code in Pre-training (Cohere)
  Including code in the pre-training data mixture, even for models not specifically designed for code, has become a common practice in LLMs p…
• The Future of International Scientific Assessments of AI’s Risks (Cohere)
  Managing the risks of artificial intelligence (AI) will require international coordination among many actors with different interests, valu…
• Bridging the Data Provenance Gap Across Text, Speech, and Video (Cohere)
  Progress in AI is driven largely by the scale and quality of training data.
• EAGER: Entropy-Aware Generation for Adaptive Inference-Time Scaling (Cohere)
  With the rise of reasoning language models and test-time scaling methods as a paradigm for improving model performance, substantial computa…
• Scalable Data Ablation Approximations for Language Models through Modular Training and Merging (Cohere)
  Training data compositions for Large Language Models (LLMs) can significantly affect their downstream performance.
• AI Exposure Scores: What they measure, what they miss, and what comes next (Cohere)
  A set of exposure scores calculated in 2023 have become a central empirical input to the future of work debate.
• Open Problems in Technical AI Governance (Cohere)
  AI progress is creating a growing range of risks and opportunities, but it is often unclear how they should be navigated.
• BigScience: A Case Study in the Social Construction of a Multilingual Large Language Model (Cohere)
  The Big Science Workshop was a value-driven initiative that spanned one and half years of interdisciplinary research and culminated in the…
• Aya Vision: Multilingual Multimodal AI Advancements (Cohere)
  May 14, 2025 Aya Vision: Multilingual Multimodal AI Advancements To address the challenges of building multimodal language models, we intro…
• On the Limitations of Compute Thresholds as a Governance Strategy (Cohere)
  At face value, this essay is about understanding a fairly esoteric governance tool called compute thresholds.
• Human Feedback is not Gold Standard (Cohere)
  Sep 28, 2023 Human Feedback is not Gold Standard Authors Tom Hosking, Phil Blunsom, Max Bartolo Abstract Human feedback has become the de f…
• SimMerge: Learning to Select Merge Operators from Similarity Signals (Cohere)
  Model merging enables multiple large language models (LLMs) to be combined into a single model while preserving performance.
• Lifting the Veil on Hyper-parameters for Value-based Deep Reinforcement Learning (Cohere)
  Successful applications of deep reinforcement learning (deep RL) combine algorithmic design and careful hyper-parameter selection.
• CALIBER: Calibrating confidence before and after reasoning in language models (Cohere)
  Reasoning language models are increasingly asked not only to answer difficult questions, but also to estimate their likelihood of success.
• SEACrowd: A Multilingual Multimodal Data Hub and Benchmark Suite for Southeast Asian Languages (Cohere)
  Authors Holy Lovenia, Rahmad Mahendra, Salsabil Maulana Akbar, Lester James V. Miranda, Jennifer Santoso, Elyanah Aco, Akhdan Fadhilah, Jon…
• Associative Memory Augmented Asynchronous Spatiotemporal Representation Learning for Event-based (Cohere)
  We propose EventFormer– a computationally efficient event-based representation learning framework for asynchronously processing event camer…
• Making, not Taking, the Best of N (Cohere)
  Obtaining high-quality generations in modern LLMs has largely been framed as a selection problem: identifying a single winning generation f…
• One Tokenizer To Rule Them All: Emergent Language Plasticity via Multilingual Tokenizers (Cohere)
  Pretraining massively multilingual Large Language Models (LLMs) for many languages at once is challenging due to limited model capacity, sc…
• Near-Optimal Distributionally Robust Reinforcement Learning with General Norms (Cohere)
  To address the challenges of sim-to-real gap and sample efficiency in reinforcement learning (RL), this work studies distributionally robus…