Papersby Glyph
Don't Mask the Environment: Observation Supervision Changes How Agents Explore Under RL research paper by Amazon, 2026
Amazon · Training and scaling · Sep 17, 2026 · 0 citations · 44 upvotes · unverified
What it shows
ActObs improves agent reinforcement learning by supervising the model to predict environment observations during supervised fine-tuning.
By Juzheng Zhang, Disha Makhija, Manoj Ghuhan Arivazhagan and 2 more · arXiv 2609.20715 · PDF
UnverifiedSummarised by a model from the abstract; not yet checked by hand.
More from Amazon
All 9ALoDLM: Adaptively Looped Diffusion Language ModelsALoDLM improves diffusion language models by using token-adaptive latent recurrence to allocate more computation to difficult tokens.Amazon1 citations80 upvotesRead →Breaking Babel: A Self-Evolving Multi-Agent System for Long-Form Subtitle TranslationSMART is a self-evolving multi-agent system that improves long-form subtitle translation by using persistent memory and dynamic routing.Amazon38 upvotesRead →Rethinking Latent Visual Reasoning: Grounding Latent Reasoning in Visual EvidenceReaLVR improves latent visual reasoning by providing explicit supervision to latent trajectories, helping models focus on relevant visual evidence.Amazon2 citations210 upvotesRead →VIDEOP2R: Video Understanding from Perception to ReasoningVideoP2R, a process-aware reinforcement fine-tuning framework, improves video reasoning and understanding by modeling perception and reasoning separately, achieving state-of-the-art results on multiple benchmarks.Amazon10 citations113 upvotesRead →Adaptive Multi-Agent Response Refinement in Conversational SystemsA multi-agent framework enhances conversational quality by refining responses through agents responsible for factuality, personalization, and coherence, outperforming existing methods on challenging datasets.Amazon3 citations42 upvotesRead →Chronos-2: From Univariate to Universal ForecastingChronos-2, a pretrained model with a group attention mechanism, achieves state-of-the-art performance in zero-shot univariate, multivariate, and covariate-informed forecasting tasks.Amazon276 citations28 upvotesRead →When Thoughts Meet Facts: Reusable Reasoning for Long-Context LMsThought templates enhance long-context language models by structuring evidence combination and guiding multi-hop inference, leading to consistent performance improvements across various benchmarks.Amazon4 citations49 upvotesRead →TaTToo: Tool-Grounded Thinking PRM for Test-Time Scaling in Tabular ReasoningTaTToo, a novel table-grounded Process Reward Model, enhances tabular reasoning by explicitly addressing table-specific operations and integrating tool-based verification, leading to significant performance improvements over existing PRMs.Amazon13 citations67 upvotesRead →
Amazon1 citations80 upvotesRead →Breaking Babel: A Self-Evolving Multi-Agent System for Long-Form Subtitle TranslationSMART is a self-evolving multi-agent system that improves long-form subtitle translation by using persistent memory and dynamic routing.Amazon38 upvotesRead →Rethinking Latent Visual Reasoning: Grounding Latent Reasoning in Visual EvidenceReaLVR improves latent visual reasoning by providing explicit supervision to latent trajectories, helping models focus on relevant visual evidence.Amazon2 citations210 upvotesRead →VIDEOP2R: Video Understanding from Perception to ReasoningVideoP2R, a process-aware reinforcement fine-tuning framework, improves video reasoning and understanding by modeling perception and reasoning separately, achieving state-of-the-art results on multiple benchmarks.Amazon10 citations113 upvotesRead →Adaptive Multi-Agent Response Refinement in Conversational SystemsA multi-agent framework enhances conversational quality by refining responses through agents responsible for factuality, personalization, and coherence, outperforming existing methods on challenging datasets.Amazon3 citations42 upvotesRead →Chronos-2: From Univariate to Universal ForecastingChronos-2, a pretrained model with a group attention mechanism, achieves state-of-the-art performance in zero-shot univariate, multivariate, and covariate-informed forecasting tasks.Amazon276 citations28 upvotesRead →When Thoughts Meet Facts: Reusable Reasoning for Long-Context LMsThought templates enhance long-context language models by structuring evidence combination and guiding multi-hop inference, leading to consistent performance improvements across various benchmarks.Amazon4 citations49 upvotesRead →TaTToo: Tool-Grounded Thinking PRM for Test-Time Scaling in Tabular ReasoningTaTToo, a novel table-grounded Process Reward Model, enhances tabular reasoning by explicitly addressing table-specific operations and integrating tool-based verification, leading to significant performance improvements over existing PRMs.Amazon13 citations67 upvotesRead →Other training and scaling papers
TopicLoRA: Low-Rank Adaptation of Large Language ModelsLoRA fine-tunes a large model by training small low-rank matrices, cutting trainable parameters by 10,000 times.
Microsoft25.6k citations66 upvotesRead →Scaling Laws for Neural Language ModelsLanguage model loss falls as a smooth power law as model size, data and compute grow.
OpenAI9,592 citations12 upvotesRead →Training Compute-Optimal Large Language ModelsChinchilla: for a fixed compute budget, train a smaller model on more data; parameters and tokens should grow together.
Google DeepMind4,035 citations13 upvotesRead →Megatron-LM: Training Multi-Billion Parameter Language Models Using Model ParallelismMegatron-LM splits each Transformer layer across GPUs to train models with billions of parameters.
NVIDIA3,387 citations5 upvotesRead →DeepSeek LLM: Scaling Open-Source Language Models with LongtermismDeepSeek LLM, an open-source language model project, develops a large dataset and employs SFT and DPO to achieve performance surpassing LLaMA-2 70B and GPT-3.5 in various benchmarks and open-ended evaluations.
DeepSeek872 citations56 upvotesRead →LoRA Learns Less and Forgets LessLoRA learns less than full fine-tuning on code and math, but forgets less of what the model already knew.
Databricks486 citations91 upvotesRead →
About this paper
- Citations
- 0, 0 influential · Semantic Scholar
- Upvotes
- 44 · Hugging Face
- Lab
- Amazon · on Companies · on Quarterly · on Paydays
Changes
ActObs improves agent reinforcement learning by supervising the model to predict environment observations during supervised fine-tuning.
Moved from Agents and evaluation to Training and scaling.
Hugging Face upvotes rose from 42 to 44.
A first count of 0 influential citations.
A first count of 0 citations.
Found by the weekly Hugging Face scan; not yet checked by hand.