Cohere · Research

Research listings from Cohere on TechiSeek. 142 public listings.

• Scale or Reason? A Compute-Equivalent Analysis of Reasoning Distillation (Cohere)
  Distilling reasoning traces from strong teacher models has become the standard recipe for building capable small language models.
• Dynamically Allocating Evaluation Effort for Model Ranking (Cohere)
  While human evaluation is the gold standard in many NLP tasks, it suffers from prohibitive costs and poor scalability.
• Findings of the WMT26 General Machine Translation Shared Task: Contrastive Dynamic Human Evaluation (Cohere)
  This paper presents the results of the General Machine Translation Task organized under the 2026 Conference on Machine Translation (WMT).
• Findings of the WMT26 Multilingual Instruction Shared Task: Small Models Are Not Yet Polyglots (Cohere)
  We present the findings of the WMT26 Multilingual Instruction Shared Task (MIST), evaluating small (≤10 B parameter) open-weight language m…
• North Small Translate: Advanced Cost-Effective Translation (Cohere CAT+) (Cohere)
  North Small Translate: Advanced Cost-Effective Translation (Cohere CAT+) We present North Small Translate, an open-weight, LLM-based machin…
• The Agentic Task Ecosystem: A Supply-Side Record of Agentic Automation (Cohere)
  Large language models increasingly use external tools to act on digital systems, but there is little systematic evidence about which kinds…
• SEATauBench: Progressively Adapting Tool-Agent-User Evaluation Into Low-Resource Southeast Asian (Cohere)
  SEATau Bench: Progressively Adapting Tool-Agent-User Evaluation Into Low-Resource Southeast Asian Languages While AI development and evalua…
• Repetition Mismatch: Why Data Mixture Experiments Don’t Scale and How to Fix Them (Cohere)
  Pre-training data mixtures are commonly tuned by running small-scale experiments and extrapolating to the target training budget.
• Contrastive ESA: Human Evaluation of Multiple Translations at Once (Cohere)
  Current human evaluation of machine translation typically assesses single outputs in isolation, a paradigm that suffers from high annotator…
• The Reality of AI and Biorisk (Cohere)
  Dec 02, 2024 The Reality of AI and Biorisk This paper provides an analysis of existing available research surrounding two AI and biorisk th…
• The Culture Funnel: You can’t align what isn’t in the data (Cohere)
  Current cultural alignment approaches focus on inference-time interventions, assuming models already contain sufficient cultural knowledge.
• BidirLM: From Text to Omnimodal Bidirectional Encoders by Adapting and Composing Causal LLMs (Cohere)
  Transforming causal generative language models into bidirectional encoders offers a powerful alternative to BERT-style architectures.
• Improving Policy Learning via Language Dynamics Distillation (Cohere)
  Recent work has shown that augmenting environments with language descriptions improves policy learning.
• RLHF Can Speak Many Languages: Unlocking Multilingual Preference Optimization for LLMs (Cohere)
  Preference optimization techniques have become a standard final stage for training state-of-art large language models (LLMs).
• Pushing Mixture of Experts to the Limit: Extremely Parameter Efficient MoE for Instruction Tuning (Cohere)
  Pushing Mixture of Experts to the Limit: Extremely Parameter Efficient MoE for Instruction Tuning Can we use Mixture of Experts (MoE) for i…
• When Personalization Meets Reality: A Multi-Faceted Analysis of Personalized Preference Learning (Cohere)
  Feb 26, 2025 When Personalization Meets Reality: A Multi-Faceted Analysis of Personalized Preference Learning We present a multi-faceted ev…
• The Presidio Recommendations on Responsible Generative AI - World Economic Forum (Cohere)
  Generative AI has the potential to transform industries and society, but responsible design and collaboration among stakeholders are crucia…
• RewardBench 2: Advancing Reward Model Evaluation (Cohere)
  Reward models are used throughout the post-training of language models to capture nuanced signals from preference data and provide a traini…
• No Need for Explanations: LLMs can implicitly learn from mistakes in-context (Cohere)
  Showing incorrect answers to Large Language Models (LLMs) is a popular strategy to improve their performance in reasoning-intensive tasks.
• The Multilingual Divide and Its Impact on Global AI Safety (Cohere)
  The Multilingual Divide and Its Impact on Global AI Safety This paper provides researchers, policymakers and governance experts with an ove…
• Here's a Free Lunch: Sanitizing Backdoored Models with Model Merge (Cohere)
  However, this openness also brings significant security risks, including backdoor attacks, where hidden malicious behaviors are triggered b…
• Diversify and Conquer: Diversity-Centric Data Selection with Iterative Refinement (Cohere)
  Finetuning large language models on instruction data is crucial for enhancing pre-trained knowledge and improving instruction-following cap…
• LLM See, LLM Do: Guiding Data Generation to Target Non-Differentiable Objectives (Cohere)
  Jul 05, 2024 LLM See, LLM Do: Guiding Data Generation to Target Non-Differentiable Objectives Our work exhaustively characterizes the impac…
• CIRCLE: A Framework for Evaluating AI from a Real-World Lens (Cohere)
  This paper proposes CIRCLE, a six-stage, lifecycle-based framework to bridge the reality gap between model-centric performance metrics and…
• Sparkles: Unlocking Chats Across Multiple Images for Multimodal Instruction-Following Models (Cohere)
  Large language models exhibit enhanced zero-shot performance on various tasks when fine-tuned with instruction-following data.
• Prioritized Training on Points that are Learnable, Worth Learning, and Not Yet Learnt (Cohere)
  Training on web-scale data can take months. But most computation and time is wasted on redundant and noisy points that are already learnt o…
• From Tools to Teammates: Evaluating LLMs in Multi-Session Coding Interactions (Cohere)
  Feb 19, 2025 From Tools to Teammates: Evaluating LLMs in Multi-Session Coding Interactions Large Language Models (LLMs) are increasingly us…
• Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought (Cohere)
  Large Language Models (LLMs) can achieve strong performance on many tasks by producing step-by-step reasoning before giving a final output,…
• Reality Check: A New Evaluation Ecosystem Is Necessary to Understand AI's Real World Effects (Cohere)
  Conventional AI evaluation approaches concentrated within the AI stack exhibit systemic limitations for exploring, navigating and resolving…
• Nexus: Specialization meets Adaptability for Efficiently Training Mixture of Experts (Cohere)
  Efficiency, specialization, and adaptability to new data distributions are qualities that are hard to combine in current Large Language Mod…
• OPERA: Automatic Offline Policy Evaluation with Re-weighted Aggregates of Multiple Estimators (Cohere)
  Offline policy evaluation (OPE) allows us to evaluate and estimate a new sequential decision-making policy's performance by leveraging hist…
• Aya 23: Open Weight Releases to Further Multilingual Progress (Cohere)
  This technical report introduces Aya 23, a family of multilingual language models.
• Soft-SVeRL: Self-Verified Reinforcement Learning with Soft Rewards (Cohere)
  Reinforcement Learning from Verifiable Rewards (RLVR) has improved language models in domains such as mathematics and code, where correctne…
• Agents Explore but Agents Ignore: LLMs Lack Environmental Curiosity (Cohere)
  LLM-based agents are assumed to integrate environmental observations into their reasoning: discovering highly relevant but unexpected infor…
• Locally Differentially Private Document Generation Using Zero Shot Prompting (Cohere)
  Numerous studies have highlighted the privacy risks associated with pretrained large language models.
• Kaleidoscope: Exams for Multilingual Vision Evaluation (Cohere)
  Apr 10, 2025 Kaleidoscope: Exams for Multilingual Vision Evaluation Kaleidoscope is a new benchmark for evaluating vision-language models a…
• Back to Basics (Cohere)
  Feb 23, 2024 Back to Basics – REINFORCE for Human Feedback in LLMs Reinforcement learning from human feedback has been widely adopted as a…
• Light bulbs have energy ratings (Cohere)
  The rising energy and environmental cost of the artificial-intelligence boom is fuelling concern.
• 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…
• 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…