2026-10-13 AI Daily | Edit Chains Address Autoregressive Shortcomings, Agent Evaluation Enters Mutual Difficulty-Setting Phase Link to heading

Today’s core shifts are concentrated across inference and evaluation: AutoIndexer trains causal models on “edit chains” to compensate for autoregression’s inability to modify outputs post-hoc; SWE-Duel introduces a self-expanding arena where agents set task difficulty for each other, driving evaluation from static to dynamic; Kairos and algebraic integer quantization offer byte-level multimodal unification and an error-free multiply-accumulate path.

📖 In-Depth Guide to This Issue’s Watch List Link to heading

Three sets of updates are worth highlighting today. First, new paradigms in inference and decoding: AutoIndexer trains causal models on “edit chains,” compensating for autoregression’s inability to perform post-hoc modifications; parallel draft context correction in speculative decoding and elastic latent recurrence optimize inference efficiency and recurrent models, respectively. Second, upgrades in Agent evaluation methodology: SWE-Duel proposes a self-expanding arena where tasks are generated during evaluation and difficulty is mutually set by agents; prompt optimization research on Deep-Research Agent measures cross-topic transfer and overfitting; perturber turns semantic-preserving perturbation into a reusable service. Third, underlying foundation models and quantization: Kairos explores byte-level multimodal unification, while algebraic integer quantization provides an error-free multiply-accumulate path, making this section well suited for readers interested in infrastructure.

🌐 X Platform AI Highlights Link to heading

Today, bird did not detect any qualifying AI topics.

💡 Influencer Insights Link to heading

Influencer insights are currently unavailable today; we recommend reading the in-depth Watch List content.

📚 Appendix: Today’s Watch List Update Source List Link to heading

Time window: Last 3 days; Covering 22 sources; Total of 32 updates

All-In Podcast (A_full) Link to heading

• Martin Shkreli: Big Pharma’s Blind Spot, Betting Against Quantum & What Prison Taught Him
  • Release time: 2026-10-12 23:49 Beijing Time
  • Summary:
    • IREN is a vertically integrated AI cloud platform providing data centers, compute, and software for AI training and inference.
    • Oracle connects data, applications, and infrastructure to transform AI into business outcomes, delivering optimized flexibility, choice, and control as AI evolves.
    • EY helps tech innovators from startups to exits to large enterprises.
    • We are committed to providing tools for everyone to reach their full potential and ensuring that all benefit from technology.
    • Keel Infrastructure possesses the power, land, and connectivity required for high-performance computing and AI operations, backed by secure energy assets and grid interconnections across North America.
  • EN Key Takeaways:
    • (0:00) Martin Shkreli joins the Besties
    • (0:55) Why quantum computing is a one-trick business and why he’s betting against it
    • (7:42) The drugs big pharma sells too cheap and his journey from hedge fund trader to pharma
    • (11:37) AI, drug discovery, and who actually wins

Stratechery by Ben Thompson (A_full) Link to heading

• Personal Day
  • Release time: 2026-10-12 18:00 Beijing Time
  • Summary:
    • $15/month or $150/year.
    • Substantive analysis of the day’s news via three weekly emails or podcasts.
    • Strategy Interviews.
    • Interviews with leading public company CEOs, private company founders, and discussions with analyst peers.
    • A twice-weekly podcast produced by John Gruber and myself: 15 minutes per episode, not a minute more or less.
  • EN Key Takeaways:
    • Personal Day — No Update

ArXiv cs.AI (B_intro+search) Link to heading

• Faithfulness Without Anchors: A Position Paper on the Structural Misalignments of Explainable AI Faithfulness Metrics for Feature Importance

  • Release time: 2026-10-12 12:00 Beijing Time
  • Summary:
    • arXiv:2610.12493v1 Announcement type: new.

• Abstract: Explainable Artificial Intelligence (XAI) tools are often presented as methods to increase trust in machine learning models by making various aspects of the algorithm more understandable to humans.

  • However, evidence accumulated over the past decade points to an overall failure of this objective, instead promoting overreliance on incorrect predictions, which can subsequently lead to algorithm aversion.
  • Rather, XAI appears to function primarily as a debugging tool, useful for obtaining anecdotal insights into, for example, features learned by the model.
  • EN Key Points:
    • arXiv:2610.12493v1 Announce Type: new
    • Abstract: Explainable Artificial Intelligence (XAI) tools have often been presented as methods to increase trust in machine learning models by making aspects of…
    • However, evidence accumulated over the last decade points to an overall failure of this objective, instead promoting overreliance on incorrect predictions, whic…
    • Rather, XAI appears to function primarily as a debugging tool, useful for obtaining anecdotal insights into, for example, features learned by the model

• EliteTable: Self-Improving Program Search for Executable Control Policies

  • Published Time: 2026-10-12 12:00 Beijing Time
  • Abstract: - arXiv:2610.12536v1 Announce Type: new.
    • Abstract: EliteTable is a self-improving program-search algorithm guided by large language models (LLMs) that uses a fixed-capacity table of high-scoring solutions to guide new proposals, edits to existing solutions, and the remixing of multiple solutions.
    • Performance was evaluated across six Gymnasium control tasks, each with ten generations and fifty candidate slots per generation.
    • Each task was evaluated across ten independently initialized searches.
  • EN Key Points:
    • arXiv:2610.12536v1 Announce Type: new
    • Abstract: EliteTable is a self-improving program-search algorithm guided by large language models (LLMs) that uses a fixed-capacity table of high-scoring soluti…
    • Performance was evaluated across six Gymnasium control tasks, each with ten generations and fifty candidate slots per generation
    • Each task was evaluated across ten independently initialized searches

• Equal-Budget Monte Carlo Tree Search versus Randomized-Restart List Scheduling for DAG Workflows: A Controlled Equal-Budget Computational Study

  • Published Time: 2026-10-12 12:00 Beijing Time
  • Abstract: - arXiv:2610.12542v1 Announce Type: new.
    • Abstract: Monte Carlo Tree Search (MCTS) is often proposed for scheduling directed acyclic graph (DAG) workflows, but most studies never perform comparisons that could refute it—by giving a simpler method the same search effort.
    • We perform comparisons under a controlled protocol.
    • All arms receive the same budget—building and evaluating the same number of candidate schedules: an MCTS scheduling method pre-registered as the primary arm; a randomized-restart list scheduler without a search tree as a control; and a textbook UCT MCTS branch as a registered follow-up branch on the same domain.
  • EN Key Points:

• arXiv:2610.12542v1 Announce Type: new

  • Abstract: Monte Carlo Tree Search (MCTS) is often proposed for scheduling directed-acyclic-graph (DAG) workflows, yet most studies never run the comparison that…
  • We run that comparison under a controlled protocol
  • All arms receive an identical budget - the same number of candidate schedules built and evaluated: an MCTS scheduling method, preregistered as the main arm; a r…

• Know What You Don’t Know: Agentic Data Synthesis by Evolving Generation Space for Closed-Loop Learning

  • Release Time: 2026-10-12 12:00 Beijing Time
  • Abstract: - arXiv:2610.12593v1 Announce Type: new.
    • Abstract: The continued improvement of large language models increasingly depends on synthetic experience, yet effective data synthesis for interactive agents requires not only executable and verifiable trajectories, but also interactions that provide learning value for the current task agent.
    • Existing methods increasingly incorporate environment feedback, data validation, or learner failures, but still face a feedback-to-decision gap: observed weaknesses do not directly specify how future synthesis should change, or whether they can be represented by the current synthesis space.
    • To address this gap, we propose EvoSyn, a learning-feedback-driven multi-agent framework that organizes synthesis around a revisable state, specifying what interaction factors can be intentionally generated and where synthesis effort should be allocated.
  • EN Key Points:
    • arXiv:2610.12593v1 Announce Type: new
    • Abstract: The continued improvement of large language models increasingly depends on synthetic experience, yet effective data synthesis for interactive agents r…
    • Existing methods increasingly incorporate environment feedback, data validation, or learner failures, but still face a feedback-to-decision gap: observed weakne…
    • To address this gap, we propose EvoSyn, a learning-feedback-driven multi-agent framework that organizes synthesis around a revisable state specifying what inter…

• SpireBench: Evaluating Long-Horizon Strategic Play of Agentic Models with Slay the Spire

  • Release Time: 2026-10-12 12:00 Beijing Time
  • Abstract: - arXiv:2610.12596v1 Announce Type: new.
    • Abstract: Agentic large language models can now handle complex tasks, but long-horizon strategic reasoning remains understudied: most agentic benchmarks are short and episodic, and performance on long-horizon tasks is difficult to disentangle into memorized knowledge, reasoning, and luck.
    • We introduce SpireBench, where models play the full commercial deck-builder Slay the Spire (StS), in which early decisions dictate long-term success.
    • Across 20 large language models (LLMs) on the game’s highest difficulty, frontier models beat heuristic baselines and match average human players, but still lag far behind experts.
  • EN Key Points:
    • arXiv:2610.12596v1 Announce Type: new

• Abstract: Agentic large language models now tackle complex tasks, yet strategic reasoning over long horizons is under-studied: most agentic benchmarks are short…

  • We introduce SpireBench, in which models play the complete commercial deck-builder Slay the Spire (StS), where early decisions decide long-term success
  • Across 20 large language models (LLMs) at the game’s highest difficulty, frontier models beat heuristic baselines and match the average human player but remain…

• Beyond Binary Verifiable Rewards: Behavioral Self-Rewarding for SWE Agents

  • Release Time: 2026-10-12 12:00 Beijing Time
  • Abstract: - arXiv:2610.12598v1 Announcement Type: new.
    • Abstract: Reinforcement learning with verifiable rewards (RLVR) provides reliable supervision for software engineering agents through executable tests, but binary rewards cannot distinguish trajectories with significantly different debugging qualities.
    • We investigate whether current policies can judge their own trajectories and use these judgments to improve task performance without a separately trained reward model. To this end, we introduce Behavioral Self-Rewarding (BSR), a reward-refinement method that reuses the policy to compare observable debugging behavior among trajectories of the same task with identical verifier outcomes.
  • EN Key Points:
    • arXiv:2610.12598v1 Announce Type: new
    • Abstract: Reinforcement learning with verifiable rewards (RLVR) provides reliable supervision for software engineering agents through executable tests, but bina…
    • We investigate whether the current policy can judge its own trajectories and use these judgments to improve task performance without a separately trained reward…
    • To this end, we introduce Behavioral Self-Rewarding (BSR), a reward-refinement method that reuses the policy to compare observable debugging behavior among traj…

• CLEAT: Learning Value-Guided Clarification for Interactive Agents

  • Release Time: 2026-10-12 12:00 Beijing Time
  • Abstract: - arXiv:2610.12655v1 Announcement Type: new.
    • Abstract: LLM agents can execute increasingly complex tasks, but effective assistance also requires aligning their actions with goals and constraints that users may not have stated.
    • We study this problem as interactive intent alignment through clarification.
    • However, clarification is only useful when agents can translate answers into better decisions.
  • EN Key Points:
    • arXiv:2610.12655v1 Announce Type: new
    • Abstract: LLM agents can execute increasingly complex tasks, but effective assistance also requires aligning their actions with goals and constraints that users…

• We study this problem as interactive intent alignment through clarification

  • Yet clarification is useful only when the agent can turn the answer into better decisions

• MagicBeanStalk: Escaping the RLVR Zero Advantage Trap with Recursive Data Synthesis

  • Publication Time: 2026-10-12 12:00 Beijing Time
  • Summary: - arXiv:2610.12659v1 Announcement Type: New.
    • Abstract: Data-centric scaling of RLVR has focused on collecting or generating verifiable training problems.
    • However, corpus size alone does not determine whether the dataset provides a useful learning signal for a target policy.
    • To guide the synthesis of effective RLVR data, we experimentally characterize dataset effectiveness as a function of its problems, the target policy, and the training rollout budget.
  • EN Key Points:
    • arXiv:2610.12659v1 Announce Type: new
    • Abstract: Data-centric scaling of RLVR has focused on collecting or generating verifiable training problems
    • However, corpus size alone does not determine whether the dataset provides a useful learning signal for a target policy
    • To guide the synthesis of effective RLVR data, we experimentally characterize dataset effectiveness as a function of its problems, the target policy, and the tr…

• Faithful and Complete LLM-Assisted Automated Construction of Hierarchical Hardware Knowledge Graphs

  • Publication Time: 2026-10-12 12:00 Beijing Time
  • Summary: - arXiv:2610.12762v1 Announcement Type: New.
    • Abstract: Hardware datasheets contain engineering specifications essential for component selection, system integration, and hardware compatibility verification.
    • However, manual inspection is tedious and error-prone because datasheets are lengthy and heterogeneous, with dense information fragmented across tables, figures, and text.
    • Although large language models (LLMs) can automate this process, naive one-shot processing omits relevant specifications even when the document fits within the model’s context window.
  • EN Key Points:
    • arXiv:2610.12762v1 Announce Type: new
    • Abstract: Hardware datasheets contain engineering specifications essential for component selection, system integration, and hardware compatibility verification
    • However, manual inspection is tedious and error-prone because datasheets are lengthy and heterogeneous, with dense information fragmented across tables, figures…
    • Although large language models (LLMs) can automate this process, na"ive one-shot processing omits relevant specifications even when the document fits within th…

• Representation Alignment from a Few Paired Data via Distilled Optimal Transport

• Release Time: 2026-10-12 12:00 Beijing Time

  • Summary: - arXiv:2610.12781v1 Announcement Type: new.
    • Abstract: Aligning two independently pretrained unimodal encoders typically demands millions of paired examples.
    • In this paper, we study the regime where only $N \sim 100$–$500$ pairs are available, while unpaired data on either side is abundant.
    • Our method expresses every representation by its similarities to the $N$ paired anchors, which places both modalities in a common coordinate system; matches the two point clouds with a single entropic optimal transport problem solved over the entire pool, whose marginal constraints counteract centrality that defeats nearest-neighbor matching; and distills the resulting plan into two lightweight encoders, thereby making the correspondence inductive.
  • EN Key Points:
    • arXiv:2610.12781v1 Announce Type: new
    • Abstract: Aligning two independently pretrained unimodal encoders typically demands millions of paired examples
    • In this paper, we study the regime where only $N \sim 100$–$500$ pairs are available, while unpaired data on either side is abundant
    • Our method expresses every representation by its similarities to the $N$ paired anchors, which places both modalities in a common coordinate system; matches the…

ArXiv cs.CL (B_intro+search) Link to heading

• AutoIndexer: Flexible-Order Decoding by Training Causal Models on Chains of Edits

  • Release Time: 2026-10-12 12:00 Beijing Time
  • Summary: - arXiv:2610.12545v1 Announcement Type: new.
    • Abstract: Autoregressive decoding is less suited to non-causally ordered data, e.g., cross-references or future dependencies, since unlike iterative human reasoning, their outputs cannot be revised post hoc.
    • Diffusion LLMs partially mitigate this limitation, but still cannot flexibly insert or remove arbitrary numbers of tokens from a pre-generated context.
    • We design AutoIndexer as a generative model with a novel training and decoding mechanism that simulates a Chain of Edits, allowing the model to flexibly modify its previous output via token insertion, replacement, and deletion.
  • EN Key Points:
    • arXiv:2610.12545v1 Announce Type: new
    • Abstract: Autoregressive decoding is less suited to non-causally ordered data, e.g., cross-references or future dependencies, since their outputs cannot be revi…
    • Diffusion LLMs partially mitigate this limitation but still cannot flexibly insert or remove arbitrary numbers of tokens from a pre-generated context
    • We design AutoIndexer as a generative model with a novel training and decoding mechanism that simulates a Chain of Edits, allowing the model to flexibly modify…

• SWE-Duel: Towards a Self-Scaling Arena for Coding Agents

  • Release Time: 2026-10-12 12:00 Beijing Time
  • Summary: - arXiv:2610.12569v1 Announcement Type: new.
    • Abstract: Static coding benchmarks are fixed at release.
    • Over time, their tasks leak into training data, and more capable agents eventually saturate them.

• We argue that a benchmark should instead aim to be self-scaling: its tasks and oracles are created at evaluation time, so no content exists in advance to leak, and its difficulty is set by participating agents rather than fixed at release.

  • EN Key Points:
    • arXiv:2610.12569v1 Announce Type: new
    • Abstract: Static coding benchmarks are fixed at release
    • Over time their tasks leak into training data, and stronger agents eventually saturate them
    • We argue that a benchmark should instead aim to be self-scaling: its tasks and oracles are created at evaluation time, so no content exists in advance to leak,…

• Contextual Correction of Parallel Draftsfor Speculative Decoding

  • Release Time: 2026-10-12 12:00 Beijing Time
  • Summary: - arXiv:2610.12570v1 Announce Type: new.
    • Abstract: Speculative decoding accelerates large language model inference by using a lightweight drafter to propose multiple tokens that are subsequently verified in parallel by a target model.
    • Recent diffusion-based drafters make this paradigm particularly attractive: deep attention-based neural backbones provide strong predictions, while drafting multiple future positions in parallel avoids the sequential cost of autoregressive generation.
    • This parallelism, however, comes at a cost.
  • EN Key Points:
    • arXiv:2610.12570v1 Announce Type: new
    • Abstract: Speculative decoding accelerates large language model inference by using a lightweight drafter to propose multiple tokens that are subsequently verifi…
    • Recent diffusion-based drafters make this paradigm particularly attractive: deep attention-based neural backbones provide strong predictions, while drafting mul…
    • This parallelism, however, comes at a cost

• CLUE: A Contrastive Learning Framework for Unsupervised Event Type Induction and Annotation

  • Release Time: 2026-10-12 12:00 Beijing Time
  • Summary: - arXiv:2610.12580v1 Announce Type: new.
    • Abstract: Event ontology is the foundational framework for defining events in a dataset.
    • Event type induction generally involves domain experts who define guidelines for creating event types and their names based on the dataset’s domain and scope.
    • However, this process is resource-intensive in terms of annotation costs and is prone to human biases.
  • EN Key Points:
    • arXiv:2610.12580v1 Announce Type: new
    • Abstract: Event ontology is the foundational framework for defining events in a dataset
    • Event type induction generally involves domain experts who define guidelines for creating event types and their names based on the dataset’s domain and scope
    • However, this process is resource-intensive in terms of annotation costs and is prone to human biases

• One Model to Loop Them All: Elastic Latent Recurrence for Language Models

  • Release Time: 2026-10-12 12:00 Beijing Time
  • Summary: - arXiv:2610.12582v1 Announcement Type: new.
    • Abstract: A latent recurrent language model repeatedly applies a shared block of layers, so a single checkpoint can trade inference compute for accuracy by changing the loop count.
    • However, at one loop, such models typically fall well short of a non-recurrent model of the same size—a one-loop tax that makes a separate non-recurrent model a better choice when latency is important.
    • We present Elastic Latent Recurrence (ELR), a method for training looped language models without this tax.
  • EN Key Points:
    • arXiv:2610.12582v1 Announce Type: new
    • Abstract: A latent recurrent language model applies a shared block of layers repeatedly, so one checkpoint can trade inference compute for accuracy by changing…
    • At one loop, however, such models typically fall well short of a non-recurrent model of the same size, a one-loop tax that makes a separate non-recurrent model…
    • We present Elastic Latent Recurrence (ELR), a recipe for training looped language models without this tax

• Reflective Prompt Optimization of a Deep-Research Agent: A Topic-Disjoint, Multi-Seed Protocol for Measuring Transferable Gains and Prompt-Level Overfitting

  • Release Time: 2026-10-12 12:00 Beijing Time
  • Summary: - arXiv:2610.12601v1 Announcement Type: new.
    • Abstract: Prompt optimization is an attractive alternative to reinforcement learning for adapting language model agents, since it is low-cost and produces human-readable artifacts.
    • However, on factual tasks, it carries a failure mode that scalar-reward reinforcement learning does not.
    • An optimizer that reads training examples can write their answers into the prompt, so in-distribution gains may be memorization rather than transferable behavior.
  • EN Key Points:
    • arXiv:2610.12601v1 Announce Type: new
    • Abstract: Prompt optimization is an attractive alternative to reinforcement learning for adapting language-model agents, since it is cheap and produces human-re…
    • On factual tasks, however, it carries a failure mode that scalar-reward reinforcement learning does not
    • An optimizer that reads the training examples can write their answers into the prompt, so an in-distribution gain may be memorization rather than transferable b…

• Self-Preservation in Language Models

  • Release Time: 2026-10-12 12:00 Beijing Time
  • Summary: - arXiv:2610.12602v1 Announcement Type: new.
    • Abstract: Continual learning (CL) seeks to adapt pre-trained models while preserving previous capabilities.
    • However, tasks provided for preservation cover only a portion of what the model knows, and forgetting may extend beyond these boundaries.

• Recent work shows that language models can supplement this limited evidence by generating useful replay without access to their original training data.

  • EN Key Points:
    • arXiv:2610.12602v1 Announce Type: new
    • Abstract: Continual learning (CL) seeks to adapt pretrained models while preserving prior capabilities
    • Yet the tasks supplied for preservation cover only part of what a model knows, while forgetting can extend beyond them
    • Recent work shows that language models can supplement this limited evidence by generating useful replay without access to their original training data

• GeopoliticalQA: Benchmarking Retrieval and Generation over Intelligence Reports

  • Publication Time: 2026-10-12 12:00 Beijing Time
  • Abstract: - arXiv:2610.12617v1 Announce Type: new.
    • Abstract: Geopolitical and strategic intelligence reports constitute a rich source of expert knowledge and have increasing value in the current political climate, but their heterogeneity and analytical complexity pose challenges to retrieval-augmented question answering.
    • We introduce a benchmark for grounded question answering over 415 open-source geopolitical and strategic intelligence reports, comprising 5,525 document-grounded question-answer pairs covering factual, reasoning, comparative, and risk-oriented questions.
    • We propose a document-aware generation pipeline that profiles reports before generating questions and validates dataset quality through expert annotation and LLM-as-a-judge agreement.
  • EN Key Points:
    • arXiv:2610.12617v1 Announce Type: new
    • Abstract: Geopolitical and strategic intelligence reports constitute a rich source of expert knowledge and have increasing value in the current political climat…
    • We introduce a benchmark for grounded question answering over 415 open-source geopolitical and strategic intelligence reports, comprising 5,525 document-grounde…
    • We propose a document-aware generation pipeline that profiles reports before generating questions and validate dataset quality through expert annotation and an…

• perturber: A Stateless Service for Semantics-Preserving Text Perturbations

  • Publication Time: 2026-10-12 12:00 Beijing Time
  • Abstract: - arXiv:2610.12667v1 Announce Type: new.
    • Abstract: Evaluating large language models (LLMs) on clean benchmarks overestimates model robustness: real-world inputs are perturbed, and these meaning-preserving variations often shift model answers.
    • Systematically studying this requires applying one perturbation at a time, repeatably and identically regardless of where they run, but existing tools couple them to benchmark stacks rather than packaging them as a reusable transformation layer.
    • We introduce perturber, a stateless service and Python library that applies only one semantics-preserving perturbation per call, drawn from a first-party catalog of 68 perturbations across 11 families.
  • EN Key Points:
    • arXiv:2610.12667v1 Announce Type: new

• Abstract: Large language model (LLM) evaluations on clean benchmarks overstate model robustness: real inputs arrive perturbed, and these meaning-preserving vari…

  • Studying this systematically requires applying perturbations one at a time, reproducibly, and identically wherever they run, yet existing tools couple them to a…
  • We present perturber, a stateless service and Python library that applies exactly one semantics-preserving perturbation per call, drawn from a first-party catal…

• GeoFitBench-LLM: A Benchmark for Assessing Large Language Models’ Capabilities in Geospatial Data Requirements Elicitation for Geo-Analytic Tasks

  • Release Time: 2026-10-12 12:00 Beijing Time
  • Abstract: - arXiv:2610.12669v1 Announce Type: new
    • Abstract: Large language models are becoming central to GeoAI systems, yet their evaluations rarely consider the elicitation of geospatial data requirements.
    • This omission is critical, as retrieval-oriented geo-analytic tasks cannot be reliably supported without first identifying the necessary datasets and constraints.
    • GeoFitBench-LLM is introduced as a benchmark for assessing whether LLMs can infer the datasets, supporting resources, and constraints required by natural language geo-analytic requests.
  • EN Key Points:
    • arXiv:2610.12669v1 Announce Type: new
    • Abstract: Large language models are becoming central to GeoAI systems, yet their evaluation rarely considers the elicitation of geospatial data requirements
    • This omission is critical as retrieval oriented geo-analytic tasks cannot be supported reliably without first identifying the necessary datasets and constraints
    • GeoFitBench-LLM is introduced as a benchmark for assessing whether LLMs can infer the datasets, the supporting resources and the constraints required by geo-ana…

ArXiv cs.LG (B_intro+search) Link to heading

• Algebraic Integer Quantization for Machine Learning Applications

  • Release Time: 2026-10-12 12:00 Beijing Time
  • Abstract: - arXiv:2610.12494v1 Announce Type: new
    • Abstract: In this paper, we propose Algebraic Integer Format (AINF) as a framework for highly configurable and precise compressed computation.
    • A key feature of AINF is that it can be directly reduced to pure integer arithmetic, thereby achieving error-free multiply-accumulate operations.
    • We develop a matrix multiplication engine that leverages this feature to achieve performance within a predictable factor of its most efficient numerical counterparts on modern hardware accelerators.
  • EN Key Points:
    • arXiv:2610.12494v1 Announce Type: new

• Abstract: In this article, we propose algebraic integer number formats (AINF) as a framework for highly configurable and accurate compressed computation

  • A key property of AINF is its straightforward reduction to purely integer arithmetic, enabling error-free multiply-accumulate operations
  • We develop a matrix multiplication engine that exploits this property, obtaining performance within a predictable factor of the most efficient numerical counter…

• Kairos: Toward a Byte-Level Foundation Model for Every Modality

  • Publication Time: 2026-10-12 12:00 Beijing Time
  • Abstract: - arXiv:2610.12495v1 Announce Type: new.
    • Abstract: Small language models trained on a single local machine leave little room for architectural trial and error.
    • Kairos combines bidirectional gated-DeltaNet linear attention and sliding-window attention over shared projections (LiZAttention2), block-wise attention-residual aggregation, sparse Mixture-of-Experts (MoE) feed-forward layers, and multi-scale byte-level codecs, trained with a unified noise-denoising objective on a single byte space shared by text, images, audio, video, LiDAR, and control streams; the report asks whether the combination is fully optimized before asking whether it is useful.
    • We provide code-derived algorithms for each component, including a cross-session recurrent-state memory gate, and then run convergence sanity checks: overfitting a single 440-byte sentence, text-only, with three seeds for codec and attention selection, and a single sweep over seven other factors.
  • EN Highlights:
    • arXiv:2610.12495v1 Announce Type: new
    • Abstract: Small language models trained on a single local machine leave little room for architectural trial and error
    • Kairos combines bidirectional gated-DeltaNet linear attention and sliding-window attention over shared projections (LiZAttention2), block-wise Attention-Residua…
    • We give code-derived algorithms for each component, including a cross-session recurrent-state memory gate, then run convergence sanity checks: overfittinga sing…

• An Input-Dependent Readout Phase for Resonant Sequence Models

  • Publication Time: 2026-10-12 12:00 Beijing Time
  • Abstract: - arXiv:2610.12496v1 Announce Type: new.
    • Abstract: Sequence models with a fixed-size state avoid the quadratic cost of attention, but face a design choice.
    • Input-dependent state transitions provide content-dependent memory at the expense of FFT-based training, whereas fixed transitions retain the O(T log T) convolution path but read memory in the same way for every input.
    • ResonatorLM, a recent model of the second class, reads a set of damped complex oscillators through a fixed learned phase per head.
  • EN Highlights:
    • arXiv:2610.12496v1 Announce Type: new
    • Abstract: Sequence models with a fixed-size state avoid the quadratic cost of attention, but face a design choice

• An input-dependent state transition gives content-dependent memory at the cost of FFT-based training, while a fixed transition keeps the O(T log T) convolutiona…

  • ResonatorLM, a recent model of the second kind, reads a bank of damped complex oscillators through a fixed learned phase per head

• CircuitTree: Composing Circuits into A Syllable Transcriber

  • Release Time: 2026-10-12 12:00 Beijing Time
  • Summary: - arXiv:2610.12520v1 Announcement Type: new.
    • Abstract: The goal of circuit discovery is to decompose a monolithic neural network into tiny modular subnetworks, each of which is interpretable and performs a recognizable subtask.
    • We develop a practical method for composing arbitrarily many circuits into a single model.
    • We propose a greedy, density-scaled composition algorithm that overcomes the limitations of previous attempts to compose circuits via naive union, then generalize it to $N$ circuits via CircuitTree, a novel data structure built on a $\Theta(N\log N)$ tree composition scheme.
  • EN Key Points:
    • arXiv:2610.12520v1 Announce Type: new
    • Abstract: The goal of circuit discovery is to decompose a monolithic neural network into tiny modular subnetworks which are interpretable in the sense that each…
    • We develop a practical method for composing arbitrarily many circuits into a single model
    • We propose a greedy, density-scaled composition algorithm that overcomes limitations of previous attempts to compose circuits by naive union, then generalize it…

• Behind-the-Meter Photovoltaic Visibility from Net Smart Meters: Deployment Regimes and Operational Insights

  • Release Time: 2026-10-12 12:00 Beijing Time
  • Summary: - arXiv:2610.12527v1 Announcement Type: new.
    • Abstract: Distribution operators need visibility of behind-the-meter photovoltaic (PV) generation, yet most residential customers are observed only through net smart meters.
    • Deep-learning-based PV disaggregation can restore this visibility, but its operational value depends on whether the model serves known customers, new customers, future periods, or new regions.
    • We quantify this dependence using four deployment regimes that separate household transfer from temporal extrapolation, applied to Australian grid smart meter data with cross-dataset adaptation on Pecan Street data.
  • EN Key Points:
    • arXiv:2610.12527v1 Announce Type: new
    • Abstract: Distribution operators need visibility of behind-the-meter photovoltaic (PV) generation, yet most residential customers are observed only through net…
    • Deep-learning-based PV disaggregation can restore this visibility, but its operational value depends on whether the model serves known customers, new customers,…

• We quantify this dependence using four deployment regimes that separate household transfer from temporal extrapolation, applied to Ausgrid smart-meter data with…

• GlucoPRISM: Protocol-Supervised Pretraining for Cross-Cohort Glucose Transfer

  • Publication Time: 2026-10-12 12:00 Beijing Time
  • Summary: - arXiv:2610.12528v1 Announcement Type: New.
    • Abstract: Continuous glucose monitoring (CGM) foundation models encode each day into a single vector that mixes who the person is, what happened that day, and how the sensor measured it, so measurement discrepancies shift as representations enter new cohorts.
    • We ask whether pretraining can place the measurement components that harm transfer into known coordinates, so that deployment can discard them with fixed projections and without target data.
    • GlucoPRISM supervises the GlucoFM backbone through response contracts and rules governing how each part of the representation should change when only the day or only the sensor changes: the 64-dimensional feature band should remain invariant to both, the 48-dimensional state band changes only across days, and the 16-dimensional sensor band tracks the device.
  • EN Key Points:
    • arXiv:2610.12528v1 Announce Type: new
    • Abstract: Continuous glucose monitoring (CGM) foundation models encode each day as one vector that mixes who the person is, what happened that day and how the s…
    • We ask whether pretraining can place the measurement component that harms transfer in known coordinates, so that deployment drops them with a fixed projection a…
    • GlucoPRISM supervises a GlucoFM backbone with response contracts, rules for how each part of the representation should change when only the day or only the sens…

• CacheScope: Exact-Compute Allocation for Training-Free Diffusion Acceleration

  • Publication Time: 2026-10-12 12:00 Beijing Time
  • Summary: - arXiv:2610.12530v1 Announcement Type: New.
    • Abstract: Training-free feature reuse accelerates diffusion transformer inference by replacing full denoising computation at selected steps with cached approximations.
    • Existing methods mainly improve the reconstruction of skipped features, but we find that once a stable first-order reconstructor is used, increasing reconstruction complexity yields limited gains, whereas the placement of exact-compute steps has a much greater impact on fidelity.
    • More importantly, absolute shallow probing changes are jointly influenced by input trajectory displacement and network response strength, and thus cannot directly reflect caching risk.
  • EN Key Points:
    • arXiv:2610.12530v1 Announce Type: new
    • Abstract: Training-free feature reuse accelerates diffusion transformer inference by replacing full denoising computation at selected steps with cached approxim…
    • Existing methods mainly improve the reconstruction of skipped features, but we find that once a stable first-order reconstructor is used, increasing reconstruct…

• More importantly, absolute shallow-probe variation is jointly affected by input-trajectory displacement and network response strength, and therefore does not di…

• Market Impact Estimation via Causal Foundation Models

  • Release Time: 2026-10-12 12:00 Beijing Time
  • Summary: - arXiv:2610.12537v1 Announce Type: new.
    • Abstract: When a trader executes a large order, the relevant question is not \emph{what will the price be?}, but \emph{what would the price have been had I not traded?}. Market impact is thus a causal quantity, yet classical models (Almgren–Chriss, propagators) regress price on flow, which is biased whenever informed flow affects price through information and trading.
    • We propose a proof-of-concept \emph{Causal Market Simulator}: a continuous-time Causal Prior-data Fitted Network (PFN) trained on synthetic market priors, where hidden information flow confounds the flow-price relationship by construction, with the confounding strength being an exact per-episode knob.
    • In this prior, the bias of the naive estimator follows a closed form that we empirically confirm (up to 68% overestimation), while PFN tracks the unbiased ground-truth parameter oracle under high confounding.
  • EN Key Points:
    • arXiv:2610.12537v1 Announce Type: new
    • Abstract: When a trader executes a large order, the relevant question is not \emph{what will the price be
    • } but \emph{what would the price have been had I not traded
    • } Market impact is thus a causal quantity, yet classical models (Almgren–Chriss, propagators) regress price on flow, which is biased whenever informed flow mov…

• Efficient Learning of Survival Analysis Models Under Outcome-Dependent Truncation and Censoring

  • Release Time: 2026-10-12 12:00 Beijing Time
  • Summary: - arXiv:2610.12538v1 Announce Type: new.
    • Abstract: Modern machine learning models are typically trained on data assumed to be representative of the predictive tasks the model will face in the future, but actual datasets violate this assumption.
    • We focus on outcome-dependent truncation, where observations appear in the training data depending on their labels.
    • Our work connects two largely distinct literatures: efficient learning under truncation and survival analysis.
  • EN Key Points:
    • arXiv:2610.12538v1 Announce Type: new
    • Abstract: Modern machine learning models are typically trained on data assumed to be representative of the predictive tasks the model will face in the future, b…
    • We focus on outcome-dependent truncation in which an observation appears in the training data depending on its label
    • Our work connects two largely distinct literatures: efficient learning under truncation and survival analysis

• Enriched Holomorphic Neural Networks for 2D and 3D Laplace Problems

  • Release Time: 2026-10-12 12:00 Beijing Time
  • Summary: - arXiv:2610.12541v1 Announce Type: new.

• Abstract: Boundary-value problems for the Laplace equation produce singular behavior near corners, edges, and junctions, which smooth approximators struggle to resolve.

  • We combine holomorphic neural networks with singular functions derived from local geometry and boundary conditions.
  • The representation is harmonic by construction, so no interior residual is needed: the network fits the smooth remainder from boundary data alone, and enrichment terms capture singular behavior.
  • EN Key Points:
    • arXiv:2610.12541v1 Announce Type: new
    • Abstract: Boundary-value problems for the Laplace equation develop singular behaviour near corners, edges and junctions that smooth approximators resolve poorly
    • We combine holomorphic neural networks with singular functions derived from the local geometry and boundary conditions
    • The representation is harmonic by construction, so no interior residual is needed: the network fits the smooth remainder from boundary data alone, and the enric…