Every paper checked — nothing hidden
All papers
Every paper checked for brief #4 of 12–19 Oct 2026 — nothing is hidden. Brief papers first.
All17,313In brief24More (on-topic)3,333Not relevant13,754Not assessed87Older352Could not confirm93Blogs with no paper266All blogs284
| Paper | Topic | Relevance | Feedback |
|---|---|---|---|
| Scaling Repeated-Data Pretraining: Batch Size, Regularization, and GPU-Time Tradeoffs This paper investigates batch size, regularization methods, and runtime tradeoffs when pretraining language models on repeated finite data. Preprint | AI agents via arxiv · 15 Oct, 06:18 | Not relevant Studies model pretraining dynamics, not LLM agents or planning. | |
| Mixed Feelings, Single Polarity The authors examine failure points when projecting complex emotion dataset labels onto single positive or negative valence polarities. Preprint | Small and efficient models via arxiv · 15 Oct, 06:18 | Not relevant Analyzes emotion label mappings rather than model efficiency. | |
| Mixed Feelings, Single Polarity The authors examine failure points when projecting complex emotion dataset labels onto single positive or negative valence polarities. Preprint | AI agents via arxiv · 15 Oct, 06:18 | Not relevant Studies NLP emotion classification, not LLM agents or planning. | |
| Mixed Feelings, Single Polarity The authors examine failure points when projecting complex emotion dataset labels onto single positive or negative valence polarities. Preprint | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18 | Not relevant Focuses on emotion polarity labeling, not retrieval systems. | |
| Mixed Feelings, Single Polarity The authors examine failure points when projecting complex emotion dataset labels onto single positive or negative valence polarities. Preprint | LLMs for code via arxiv · 15 Oct, 06:18 | Not relevant Investigates sentiment and emotion datasets, not code generation. | |
| Mixed Feelings, Single Polarity The authors examine failure points when projecting complex emotion dataset labels onto single positive or negative valence polarities. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18 | Not relevant Examines classical sentiment dataset annotations rather than LLM benchmarks. | |
| Mixed Feelings, Single Polarity The authors examine failure points when projecting complex emotion dataset labels onto single positive or negative valence polarities. Preprint | Multimodal models via arxiv · 15 Oct, 06:18 | Not relevant Studies text sentiment analysis, not multimodal systems. | |
| Mixed Feelings, Single Polarity The authors examine failure points when projecting complex emotion dataset labels onto single positive or negative valence polarities. Preprint | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | Not relevant Studies emotion classification datasets, not deployment safety. | |
| PAIR: Memory-Efficient and Fast Inference with Reading--Reasoning Fusion PAIR splits context processing and answer reasoning between two different-sized models, fusing predictions at the logit level for efficient inference. Preprint | LLMs for code via arxiv · 15 Oct, 06:18 | Not relevant Applies to context-based QA rather than software engineering. | |
| PAIR: Memory-Efficient and Fast Inference with Reading--Reasoning Fusion PAIR splits context processing and answer reasoning between two different-sized models, fusing predictions at the logit level for efficient inference. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18 | Not relevant Presents an inference architecture rather than evaluation benchmarks. | |
| PAIR: Memory-Efficient and Fast Inference with Reading--Reasoning Fusion PAIR splits context processing and answer reasoning between two different-sized models, fusing predictions at the logit level for efficient inference. Preprint | AI agents via arxiv · 15 Oct, 06:18 | Not relevant Focuses on split-model inference rather than multi-step agent planning. | |
| PAIR: Memory-Efficient and Fast Inference with Reading--Reasoning Fusion PAIR splits context processing and answer reasoning between two different-sized models, fusing predictions at the logit level for efficient inference. Preprint | Multimodal models via arxiv · 15 Oct, 06:18 | Not relevant Focuses on language model context processing without multimodal inputs. | |
| PAIR: Memory-Efficient and Fast Inference with Reading--Reasoning Fusion PAIR splits context processing and answer reasoning between two different-sized models, fusing predictions at the logit level for efficient inference. Preprint | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18 | Not relevant Studies context inference fusion rather than external retrieval systems. | |
| PAIR: Memory-Efficient and Fast Inference with Reading--Reasoning Fusion PAIR splits context processing and answer reasoning between two different-sized models, fusing predictions at the logit level for efficient inference. Preprint | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | Not relevant Focuses on inference latency and memory rather than safety. | |
| Improving the Evaluation of Biomedical Question Answering with Few Human Preference Signals This paper uses limited human preference signals to calibrate automated evaluation metrics for biomedical retrieval-augmented question answering. Preprint | LLMs for code via arxiv · 15 Oct, 06:18 | Not relevant Evaluates biomedical QA, not software engineering. | |
| Improving the Evaluation of Biomedical Question Answering with Few Human Preference Signals This paper uses limited human preference signals to calibrate automated evaluation metrics for biomedical retrieval-augmented question answering. Preprint | Multimodal models via arxiv · 15 Oct, 06:18 | Not relevant Focuses on text biomedical QA, not multimodal data. | |
| Improving the Evaluation of Biomedical Question Answering with Few Human Preference Signals This paper uses limited human preference signals to calibrate automated evaluation metrics for biomedical retrieval-augmented question answering. Preprint | AI agents via arxiv · 15 Oct, 06:18 | Not relevant Focuses on evaluation metrics for QA rather than agent planning. | |
| Improving the Evaluation of Biomedical Question Answering with Few Human Preference Signals This paper uses limited human preference signals to calibrate automated evaluation metrics for biomedical retrieval-augmented question answering. Preprint | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | Not relevant Addresses evaluation metric alignment rather than AI safety guardrails. | |
| Improving the Evaluation of Biomedical Question Answering with Few Human Preference Signals This paper uses limited human preference signals to calibrate automated evaluation metrics for biomedical retrieval-augmented question answering. Preprint | Small and efficient models via arxiv · 15 Oct, 06:18 | Not relevant Focuses on evaluation methodology, not model compression or efficiency. | |
| Beyond the greenwash: do LLMs' ecologically sound decisions hold up under pressure? This paper measures whether ecological alignment in large language models survives under pressure using a calibrated pretraining ladder. Preprint | LLMs for code via arxiv · 15 Oct, 06:18 | Not relevant Studies ecological bias and decision-making, not software engineering. | |
| Beyond the greenwash: do LLMs' ecologically sound decisions hold up under pressure? This paper measures whether ecological alignment in large language models survives under pressure using a calibrated pretraining ladder. Preprint | Small and efficient models via arxiv · 15 Oct, 06:18 | Not relevant Studies pretraining alignment and evaluation rather than model compression. | |
| Beyond the greenwash: do LLMs' ecologically sound decisions hold up under pressure? This paper measures whether ecological alignment in large language models survives under pressure using a calibrated pretraining ladder. Preprint | AI agents via arxiv · 15 Oct, 06:18 | Not relevant Focuses on behavioral alignment testing rather than agent workflows. | |
| Beyond the greenwash: do LLMs' ecologically sound decisions hold up under pressure? This paper measures whether ecological alignment in large language models survives under pressure using a calibrated pretraining ladder. Preprint | Multimodal models via arxiv · 15 Oct, 06:18 | Not relevant Evaluates language models without multimodal components. | |
| Beyond the greenwash: do LLMs' ecologically sound decisions hold up under pressure? This paper measures whether ecological alignment in large language models survives under pressure using a calibrated pretraining ladder. Preprint | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18 | Not relevant Investigates model disposition under pressure, not retrieval systems. | |
| Scaling Multi-Agent Intelligence with Substrate-Centric Coordination ScaleMAS coordinates multi-agent systems using an asynchronous shared datastore with limited search to improve scaling and fault tolerance. Preprint | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18 | Not relevant Focuses on multi-agent messaging substrates rather than external knowledge RAG. | |
| Scaling Multi-Agent Intelligence with Substrate-Centric Coordination ScaleMAS coordinates multi-agent systems using an asynchronous shared datastore with limited search to improve scaling and fault tolerance. Preprint | Multimodal models via arxiv · 15 Oct, 06:18 | Not relevant Deals with multi-agent coordination without multimodal data. | |
| Scaling Multi-Agent Intelligence with Substrate-Centric Coordination ScaleMAS coordinates multi-agent systems using an asynchronous shared datastore with limited search to improve scaling and fault tolerance. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18 | Not relevant Proposes a multi-agent framework rather than evaluation benchmarks. | |
| Scaling Multi-Agent Intelligence with Substrate-Centric Coordination ScaleMAS coordinates multi-agent systems using an asynchronous shared datastore with limited search to improve scaling and fault tolerance. Preprint | LLMs for code via arxiv · 15 Oct, 06:18 | Not relevant Addresses general multi-agent coordination rather than software engineering. | |
| Scaling Multi-Agent Intelligence with Substrate-Centric Coordination ScaleMAS coordinates multi-agent systems using an asynchronous shared datastore with limited search to improve scaling and fault tolerance. Preprint | Small and efficient models via arxiv · 15 Oct, 06:18 | Not relevant Focuses on agent coordination mechanisms rather than model compression. | |
| Hashrope: A Persistent Data Structure for Mechanisms of LLM Context Management This paper presents Hashrope, a unified persistent data structure supporting incremental edits, branching, and cache reuse for language model serving. Preprint | AI agents via arxiv · 15 Oct, 06:18 | Not relevant Covers data structures for serving stacks, not agent planning. | |
| Hashrope: A Persistent Data Structure for Mechanisms of LLM Context Management This paper presents Hashrope, a unified persistent data structure supporting incremental edits, branching, and cache reuse for language model serving. Preprint | LLMs for code via arxiv · 15 Oct, 06:18 | Not relevant Focuses on serving data structures rather than code assistants. | |
| Hashrope: A Persistent Data Structure for Mechanisms of LLM Context Management This paper presents Hashrope, a unified persistent data structure supporting incremental edits, branching, and cache reuse for language model serving. Preprint | Multimodal models via arxiv · 15 Oct, 06:18 | Not relevant Addresses text context serving engines, not multimodal systems. | |
| Hashrope: A Persistent Data Structure for Mechanisms of LLM Context Management This paper presents Hashrope, a unified persistent data structure supporting incremental edits, branching, and cache reuse for language model serving. Preprint | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | Not relevant Focuses on inference serving mechanics rather than deployment safety. | |
| Hashrope: A Persistent Data Structure for Mechanisms of LLM Context Management This paper presents Hashrope, a unified persistent data structure supporting incremental edits, branching, and cache reuse for language model serving. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18 | Not relevant Introduces a context management data structure, not an evaluation benchmark. | |
| Hashrope: A Persistent Data Structure for Mechanisms of LLM Context Management This paper presents Hashrope, a unified persistent data structure supporting incremental edits, branching, and cache reuse for language model serving. Preprint | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18 | Not relevant Focuses on KV cache and context management, not RAG retrieval. | |
| Think Before You Speak: Enabling Training-Free Latent Communication Between Language Models This paper presents CacheFuse, a training-free technique enabling heterogeneous language models to communicate directly by sharing key-value cache states. Preprint | Multimodal models via arxiv · 15 Oct, 06:18 | Not relevant Focuses on language model latent communication, not multimodal models. | |
| Think Before You Speak: Enabling Training-Free Latent Communication Between Language Models This paper presents CacheFuse, a training-free technique enabling heterogeneous language models to communicate directly by sharing key-value cache states. Preprint | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18 | Not relevant Studies inter-model KV-cache sharing rather than document retrieval or RAG. | |
| Think Before You Speak: Enabling Training-Free Latent Communication Between Language Models This paper presents CacheFuse, a training-free technique enabling heterogeneous language models to communicate directly by sharing key-value cache states. Preprint | LLMs for code via arxiv · 15 Oct, 06:18 | Not relevant Focuses on general inter-model communication rather than software engineering. | |
| Think Before You Speak: Enabling Training-Free Latent Communication Between Language Models This paper presents CacheFuse, a training-free technique enabling heterogeneous language models to communicate directly by sharing key-value cache states. Preprint | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | Not relevant Focuses on inter-model communication speed, not deployment safety. | |
| Think Before You Speak: Enabling Training-Free Latent Communication Between Language Models This paper presents CacheFuse, a training-free technique enabling heterogeneous language models to communicate directly by sharing key-value cache states. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18 | Not relevant Proposes a collaboration method rather than benchmark evaluation methodologies. | |
| DeepThink: A Conceptual Model for Knowledge-Addressable Recurrent Language Modeling This paper proposes DeepThink, a conceptual architecture that decouples knowledge capacity, activation width, and recurrent depth in language models. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18 | Not relevant Introduces a conceptual model architecture, not evaluation benchmarks. | |
| DeepThink: A Conceptual Model for Knowledge-Addressable Recurrent Language Modeling This paper proposes DeepThink, a conceptual architecture that decouples knowledge capacity, activation width, and recurrent depth in language models. Preprint | AI agents via arxiv · 15 Oct, 06:18 | Not relevant Proposes a recurrent architecture rather than multi-step agent systems. | |
| DeepThink: A Conceptual Model for Knowledge-Addressable Recurrent Language Modeling This paper proposes DeepThink, a conceptual architecture that decouples knowledge capacity, activation width, and recurrent depth in language models. Preprint | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | Not relevant Proposes a recurrent model architecture rather than deployment safety mechanisms. | |
| DeepThink: A Conceptual Model for Knowledge-Addressable Recurrent Language Modeling This paper proposes DeepThink, a conceptual architecture that decouples knowledge capacity, activation width, and recurrent depth in language models. Preprint | LLMs for code via arxiv · 15 Oct, 06:18 | Not relevant Presents a conceptual language model architecture, not code generation. | |
| DeepThink: A Conceptual Model for Knowledge-Addressable Recurrent Language Modeling This paper proposes DeepThink, a conceptual architecture that decouples knowledge capacity, activation width, and recurrent depth in language models. Preprint | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18 | Not relevant Focuses on internal parameter-space knowledge routing, not RAG. | |
| DeepThink: A Conceptual Model for Knowledge-Addressable Recurrent Language Modeling This paper proposes DeepThink, a conceptual architecture that decouples knowledge capacity, activation width, and recurrent depth in language models. Preprint | Multimodal models via arxiv · 15 Oct, 06:18 | Not relevant Focuses entirely on language modeling without multimodal inputs. | |
| Lüshi-1840: Building and Probing a Knowledge Cutoff for a Classical Chinese Language Model This paper introduces Lüshi-1840, a pre-modern Chinese language model trained on historical texts to evaluate knowledge cutoffs and hindsight bias. Preprint | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | Not relevant Explores historical corpus cutoff rather than deployment safety or jailbreaks. | |
| Lüshi-1840: Building and Probing a Knowledge Cutoff for a Classical Chinese Language Model This paper introduces Lüshi-1840, a pre-modern Chinese language model trained on historical texts to evaluate knowledge cutoffs and hindsight bias. Preprint | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18 | Not relevant Examines pretraining cutoff corpora rather than retrieval-augmented generation. | |
| Lüshi-1840: Building and Probing a Knowledge Cutoff for a Classical Chinese Language Model This paper introduces Lüshi-1840, a pre-modern Chinese language model trained on historical texts to evaluate knowledge cutoffs and hindsight bias. Preprint | LLMs for code via arxiv · 15 Oct, 06:18 | Not relevant Investigates classical Chinese text modeling, not code. | |
| Lüshi-1840: Building and Probing a Knowledge Cutoff for a Classical Chinese Language Model This paper introduces Lüshi-1840, a pre-modern Chinese language model trained on historical texts to evaluate knowledge cutoffs and hindsight bias. Preprint | Multimodal models via arxiv · 15 Oct, 06:18 | Not relevant Deals with historical Chinese text, not multimodal models. | |
| Lüshi-1840: Building and Probing a Knowledge Cutoff for a Classical Chinese Language Model This paper introduces Lüshi-1840, a pre-modern Chinese language model trained on historical texts to evaluate knowledge cutoffs and hindsight bias. Preprint | AI agents via arxiv · 15 Oct, 06:18 | Not relevant Focuses on historical language pretraining, not agent workflows. | |
| Lüshi-1840: Building and Probing a Knowledge Cutoff for a Classical Chinese Language Model This paper introduces Lüshi-1840, a pre-modern Chinese language model trained on historical texts to evaluate knowledge cutoffs and hindsight bias. Preprint | Small and efficient models via arxiv · 15 Oct, 06:18 | Not relevant Focuses on temporal dataset curation rather than model compression. | |
| Error Correction in Trained TTT Layers This paper analyzes prediction-derived error updates and test-time gradient objectives in trained test-time-training layers. Preprint | Retrieval-augmented generation (RAG) via arxiv · 15 Oct, 06:18 | Not relevant Examines internal memory mechanisms, not external knowledge retrieval systems. | |
| Error Correction in Trained TTT Layers This paper analyzes prediction-derived error updates and test-time gradient objectives in trained test-time-training layers. Preprint | LLM evaluation and benchmarks via arxiv · 15 Oct, 06:18 | Not relevant Studies gradient mechanics in TTT layers rather than evaluation benchmarks. | |
| Error Correction in Trained TTT Layers This paper analyzes prediction-derived error updates and test-time gradient objectives in trained test-time-training layers. Preprint | AI agents via arxiv · 15 Oct, 06:18 | Not relevant Studies test-time training layer mechanics, not agent tool use or planning. | |
| Error Correction in Trained TTT Layers This paper analyzes prediction-derived error updates and test-time gradient objectives in trained test-time-training layers. Preprint | LLMs for code via arxiv · 15 Oct, 06:18 | Not relevant Analyzes model architecture updates rather than code generation. | |
| Error Correction in Trained TTT Layers This paper analyzes prediction-derived error updates and test-time gradient objectives in trained test-time-training layers. Preprint | Multimodal models via arxiv · 15 Oct, 06:18 | Not relevant Focuses on sequence model layers without multimodal integration. | |
| Error Correction in Trained TTT Layers This paper analyzes prediction-derived error updates and test-time gradient objectives in trained test-time-training layers. Preprint | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | Not relevant Analyzes architectural inner-loop updates rather than safety or reliability. | |
| Deep Speech Representations Track Word Timing along a Low-Dimensional Circular Trajectory This paper demonstrates that deep speech representations track word timing along low-dimensional circular trajectories across multiple model architectures. Preprint | Multimodal models via arxiv · 15 Oct, 06:18 | Not relevant Analyzes speech-only representation geometry rather than multimodal cross-modality reasoning. | |
| Deep Speech Representations Track Word Timing along a Low-Dimensional Circular Trajectory This paper demonstrates that deep speech representations track word timing along low-dimensional circular trajectories across multiple model architectures. Preprint | AI safety and reliability in deployment via arxiv · 15 Oct, 06:18 | Not relevant Studies acoustic representational geometry, not safety or guardrails. |
7,381–7,440 of 13,754← PreviousNext →