System translated (Gemini)

🤖 AI 速览

Today’s focus is not on single-point model refreshes, but on capabilities entering a more rigorous validation and delivery phase. Google DeepMind is advancing double-blind evaluations and more controllable generation capabilities, while products like speech transcription and video generation …
📋 文章元数据
发布时间
2026-08-28
类型
ai-daily
字数
6937
阅读时长
33 min

2026-08-28 AI Daily | From Usable to Verifiable: Double-Blind Evaluation Arrives as Voice and Video Capabilities Advance Link to heading

Today’s focus isn’t on isolated model updates, but on capabilities entering a more rigorous phase of validation and delivery. Google DeepMind is advancing double-blind evaluations and more controllable generation capabilities, while products for speech-to-text and video generation continue to move towards low latency, long context, and production workflows. Another key thread is the re-evaluation of benchmark reliability, as many metrics are beginning to reveal the biases and distortions of the tests themselves.

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

There are three main themes to follow today. The first is “Model Controllability and Post-Training”: Google DeepMind’s Gemini Omni 1.1 Flash, double-blind AI evaluations, and several papers on unsupervised post-training, activation steering, and the effects of fine-tuning all point to the same issue—models are growing more powerful, but the boundaries of what is truly controllable and verifiable are still being redrawn. The second is “Benchmark Reliability”: From dialect bias and semantic reply consistency to the “imperfective paradox” distorting benchmarks, multiple studies today warn that for many seemingly stable metrics, the tests themselves may be the first to break. The third is “Structured Task Implementation”: ESQ-Bench, DataKernelBench, and new frameworks for evaluating memory/RAG show that enterprise-level NL2SQL, database optimization, and retrieval systems are now undergoing more rigorous real-world scenario testing.

🌐 AI Hot Topics on X Link to heading

Topic 1: Cursor Launches Scratch-to-Deploy Web App Builder Link to heading

  • Category: AI · News
  • Summary: Trending Time: , Related Posts: 203
  • What it is: Cursor has launched a web app builder that can build from scratch and deploy directly, advancing AI programming from prototype creation to live delivery.
  • Why it matters: This indicates that AI programming tools are shifting from ‘code assistance’ to ’end-to-end application delivery,’ which directly impacts development workflows, product iteration speed, and the commercialization of AI-native applications.
  • Discussion summary: The discussion on X focuses on two main points: first, whether this Scratch-to-Deploy approach can genuinely lower the barrier for non-engineering users to create functional applications; and second, whether it represents a new paradigm or merely an integration of capabilities compared to existing low-code platforms, IDEs, and AI coding assistants.

Topic 2: Tech Giants Urge Urgent AI Cyber Defense Action Link to heading

  • Category: AI · News
  • Summary: Trending Time: 6 hours ago, Related Posts: 8200
  • What it is: Several major tech companies are calling for more urgent action on AI cyber defense measures to address the risk of generative AI being used for attacks and the resulting imbalance in security.
  • Why it matters: This is important because AI is simultaneously amplifying both offensive and defensive capabilities. If protection systems don’t keep pace, models, data, and infrastructure could become targets for larger-scale cyber threats.
  • Discussion summary: The discussion on X centers on two points: whether businesses and governments have underestimated the security risks posed by AI, and whether the focus should be on industry self-regulation, technical standards, or stricter regulations to drive the implementation of AI cyber defenses.

Topic 3: Google Launches Gemini 3.5 Transcribe for Precise Speech-to-Text Link to heading

  • Category: AI · News
  • Summary: Trending Time: 1 day ago, Related Posts: 5500
  • What it is: Google has released Gemini 3.5 Transcribe for high-precision speech-to-text. It supports automatic recognition of over 85 languages, removal of filler words, multi-speaker differentiation, and offers both offline and low-latency live streaming modes.
  • Why it matters: This signifies that speech recognition is evolving from a general capability into an infrastructure-level service that can be directly embedded into production workflows. It has a direct impact on meeting minutes, customer service, media transcription, and multilingual applications, and reflects the competition in AI productization shifting towards real-time performance, accuracy, and ease of developer integration.
  • Discussion summary: The discussion on X focuses on several points: whether its multilingual and low-latency performance is sufficient to compete with existing transcription solutions, the value of its 85-language support and custom vocabularies for industry-specific scenarios, and whether Google is using this to further embed voice capabilities into the Gemini ecosystem and developer tools.

Topic 4: Google Launches Gemini Omni 1.1 Flash for Advanced Video Creation Link to heading

  • Category: AI · News
  • Summary: Trending Time: 7 hours ago, Related Posts: 3000
  • What it is: Google has released Gemini Omni 1.1 Flash for more advanced video creation, adding new capabilities like scene extension and the ability to analyze up to 10 seconds of footage to maintain narrative coherence.
  • Why it matters: This shows that generative video models are moving from single-clip generation towards stronger contextual understanding and continuity control, which directly impacts the usability, cost, and production efficiency of AI video creation.
  • Discussion Summary: Discussions on X are focused on whether it truly improves video narrative consistency, its actual performance compared to existing video generation models, and how such capabilities will impact content creation workflows, copyright, and authenticity issues.

Topic 5: Salesforce Beats Earnings Expectations with Claude AI Integration Link to heading

  • Category: AI · News
  • Overview: Trending Time: 1 day ago, Related Posts: 8100
  • What Happened: Salesforce announced earnings that exceeded market expectations, highlighting its Claude AI integration as a key performance driver.
  • Why It Matters: This indicates that generative AI is moving from proof-of-concept to driving actual revenue and product differentiation in enterprise software, signaling progress in AI’s commercialization path and enterprise-level adoption.
  • Discussion Summary: Discussions on X are centered on whether the AI integration is genuinely driving Salesforce’s growth, Claude’s competitiveness in the enterprise sector, and the impact of such collaborations on revenue, profit margins, and dependence on Anthropic.

Topic 6: Nvidia Acquires Hugging Face for $12.9 Billion in Major AI Deal Link to heading

  • Category: AI · News
  • Overview: Trending Time: 19 hours ago, Related Posts: 22000
  • What Happened: A trending report on X claims that Nvidia has acquired Hugging Face for $12.9 billion, a deal that has attracted widespread attention in the AI community.
  • Why It Matters: Hugging Face is a critical gateway to the open-source model and tool ecosystem. If acquired by Nvidia, it could reshape the power structure of AI infrastructure, model distribution, and the developer ecosystem.
  • Discussion Summary: The discussion is focused on the impact of this deal on open-source neutrality, whether Nvidia will further strengthen its control over the full AI stack, and if this will change the choices of the model community and enterprise users.

Topic 7: Neil Movva Breaks Down AI Inference Economics on Invest Like the Best Link to heading

  • Category: AI · News
  • Overview: Trending Time: 2 days ago, Related Posts: 4100
  • What Happened: On the “Invest Like the Best” podcast, Neil Movva discussed and broke down the economic model of AI inference, focusing on compute costs, pricing, and commercialization paths.
  • Why It Matters: Inference costs directly determine the gross margins, scaling speed, and competitive landscape of AI products. Therefore, this type of analysis influences the business decisions of model vendors, cloud service providers, and application-layer companies.
  • Discussion Summary: Discussions on X are centered on whether inference costs will decline rapidly, who will gain an advantage in compute and infrastructure, and whether AI companies can build sustainable revenue models amid high compute consumption.

Topic 8: Tesla Adds 79 Model Ys to Texas Robotaxi Fleet in One Day Link to heading

  • Category: AI · News
  • Overview: Trending Time:, Related Posts: 823
  • What Happened: Tesla reportedly added 79 Model Y vehicles to its Texas Robotaxi fleet in a single day for the expansion of its autonomous ride-hailing service.
  • Why It Matters: This suggests Tesla is accelerating the conversion of its mass-produced cars into an operational autonomous fleet, which is crucial for the real-world application, large-scale deployment, and commercial validation of AI in transportation.
  • Discussion Summary: Discussions on X are focused on whether this marks a substantial expansion phase for the Robotaxi project, and whether the vehicles’ autonomous capabilities, regulatory compliance, operational range, and data feedback are sufficient to support Tesla’s narrative. The main point of contention is whether this represents a genuine commercial breakthrough or merely a limited fleet expansion and marketing effort.

Today’s AI Public Opinion Summary on X Link to heading

Today’s main narrative is consistent: AI is shifting from “demo-ready” to “delivery-ready.” Whether it’s Cursor’s zero-to-deployment, Google’s speech-to-text and video generation, or Salesforce’s enterprise revenue validation, the focus is on usability, integration, and commercialization. The clear consensus is that the deciding factor for success is no longer standalone model capabilities, but rather the ability to integrate into workflows, lower the barrier to entry, and create a viable business model based on inference costs and product revenue. Disagreements primarily fall into two categories: first, whether these announcements represent a paradigm shift or simply a repackaging of existing IDE, low-code, cloud, and model capabilities; and second, whether news like the Tesla Robotaxi fleet and the Nvidia-Hugging Face acquisition signifies substantial expansion or is driven more by capital and narrative. Potential risks are also repeatedly mentioned: the imbalance in AI-driven cyber offense and defense will expose models, data, and infrastructure to more frequent and larger-scale threats; enhanced video and transcription capabilities will amplify issues of copyright, authenticity, and content misuse; and if infrastructure and distribution channels continue to consolidate among a few giants, open-source neutrality, developer choice, and market competition will be under pressure. Overall, today’s debate is not about if AI will be implemented, but about who can turn it into a real business with sustainable costs, within regulatory boundaries, and with control over the ecosystem.

💡 Influencer Insights Link to heading

Influencer insights are unavailable today. We recommend reading the in-depth content from the Watch List.

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

Time window: last 3 days; covers 22 sources; 34 updates in total.

OpenAI Blog (A_full) Link to heading

  • Better answers, broader thinking: What students gain from ChatGPT and critical-thinking training

    • Published: 2026-08-27 17:00 Beijing Time
    • Summary: [Translation Pending] - What happens when students use ChatGPT on a real-world assignment?
      • Do quality improvements come at the expense of originality?
      • A new experiment from researchers at Bocconi University, in collaboration with OpenAI Economic Research, found distinct and complementary effects from ChatGPT access and critical-thinking training.
      • Access to ChatGPT improved the quality and coherence of students’ work, while an exercise in causal reasoning—a form of critical thinking—led students to generate more unique ideas.
      • Students who received both ChatGPT access and the training showed both effects.
    • EN Key points:
      • A randomized study of more than 1,000 students examines ChatGPT, critical thinking, originality, and student performance on a real-world university assignment.

  • Expanding OpenAI’s presence in Brazil

    • Published: 2026-08-27 11:00 Beijing Time
    • Summary: [Translation Pending] - We’re excited to expand our work in Brazil with the launch of our commercial operations.
      • Based in São Paulo, our local team will work with Brazilian businesses, developers, researchers, and public institutions to help translate the country’s rapid adoption of AI into economic growth and meaningful progress.
      • Brazil is one of ChatGPT’s three largest markets by weekly active users.
      • The number of users in the country has nearly doubled over the past year, and people in Brazil now send approximately 215 million messages to ChatGPT each day.

      • “The most exciting part isn’t just the scale of adoption.

    • EN Key points:

  • OpenAI is expanding its presence in Brazil, deepening engagement with developers, businesses, and communities to support AI adoption across the country.

Google DeepMind Blog (A_full) Link to heading

  • Gemini Omni 1.1 Flash lets you build with more control

    • Published at: 2026-08-28 00:11 Beijing Time
    • Summary: [To be translated] - Gemini Omni 1.1 Flash lets you build with more control.
      • This piece from Google DeepMind Blog explains how Gemini Omni 1.1 Flash lets you build with more control shapes the broader AI and infrastructure landscape.
      • It also surfaces practical implications for founders, operators, and investors following Gemini Omni 1.1 Flash lets you build with more control.
    • EN Highlights:
      • Gemini Omni 1.1 Flash lets you build with more control

  • Piloting the world’s first double-blind AI evaluations

    • Published at: 2026-08-27 20:59 Beijing Time
    • Summary: [To be translated] - Piloting the world’s first double-blind AI evaluations.
      • This piece from Google DeepMind Blog explains how Piloting the world’s first double-blind AI evaluations shapes the broader AI and infrastructure landscape.
      • It also surfaces practical implications for founders, operators, and investors following Piloting the world’s first double-blind AI evaluations.
    • EN Highlights:
      • Piloting the world’s first double-blind AI evaluations

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

  • RENDER: Controlling Reader-Facing Evidence in LLM Memory Evaluation

    • Published at: 2026-08-27 12:00 Beijing Time
    • Summary: [To be translated] - arXiv:2608.23568v1 Announce Type: new.
      • Abstract: Memory and RAG evaluations often treat the answering model’s input as an implementation detail, even though systems may render the same history as a memory entry, summary, typed record, or raw excerpt.
      • We introduce RENDER, a benchmark control that fixes the conversation while varying the reader-facing artifact.

  • RENDER combines a five-level packet ladder, localizing when answer-bearing content enters the input, with deterministic templates approximating ChatGPT-style entries, LangChain summaries, MemGPT-style typed records, and raw conversation.

    • EN Key Points:
      • arXiv:2608.23568v1 Announce Type: new
      • Abstract: Memory and RAG evaluations often treat the answering model’s input as an implementation detail, even though systems may render the same history as a m…
      • We introduce RENDER, a benchmark control that fixes the conversation while varying the reader-facing artifact
      • RENDER combines a five-level packet ladder, localizing when answer-bearing content enters the input, with deterministic templates approximating ChatGPT-style en…

  • ESQ-Bench: A Multi-Tier Enterprise Oracle Benchmark for Evaluating NL2SQL Dialect Generalization and Silent Semantic Divergence

    • Publication Time: 2026-08-27 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2608.23569v1 Announce Type: new.
      • Abstract: State-of-the-art Natural Language to SQL (NL2SQL) models report execution accuracy exceeding 89 percent on established benchmarks such as Spider and BIRD.
      • However, these benchmarks rely on simplified academic schemas and open-source SQL dialects that do not reflect the complexity of enterprise database environments.
      • We introduce ESQ-Bench, an Oracle-first NL2SQL benchmark with systematic complexity tiers and silent-divergence evaluation across three enterprise schema complexity tiers.
    • EN Key Points:
      • arXiv:2608.23569v1 Announce Type: new
      • Abstract: State-of-the-art Natural Language to SQL (NL2SQL) models report execution accuracy exceeding 89 percent on established benchmarks such as Spider and B…
      • However, these benchmarks rely on simplified academic schemas and open-source SQL dialects that do not reflect the complexity of enterprise database environment…

  • We introduce ESQ-Bench, an Oracle-first NL2SQL benchmark with systematic complexity tiers and silent-divergence evaluation across three enterprise schema comple…

  • LLM Agents Perform Controlled Experiments Using Simulation Models

    • Publication Time: 2026-08-27 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2608.23622v1 Announce Type: new.
      • Abstract: Large language models (LLMs) have shown strong capabilities in reasoning, planning, and tool use, but many scientific and engineering tasks require more than plausible text and code generation.
      • They require understanding how a system responds to intervention, which in practice depends on controlled experimentation.
      • In this work, we propose a multi-agent framework that enables LLM agents to conduct controlled experiments with scientific simulation models for pharmaceutical process design.
    • EN Key Points:
      • arXiv:2608.23622v1 Announce Type: new
      • Abstract: Large language models (LLMs) have shown strong capabilities in reasoning, planning, and tool use, but many scientific and engineering tasks require mo…
      • They require understanding how a system responds to intervention, which in practice depends on controlled experimentation
      • In this work, we propose a multi-agent framework that enables LLM agents to conduct controlled experiments with scientific simulation models for pharmaceutical…

  • A survey detection channel overrides the pixels in an astronomical foundation model, and biases tomographic mean redshifts

    • Publication Time: 2026-08-27 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2608.23626v1 Announce Type: new.
      • Abstract: Foundation models for astronomy are trained on survey pixels together with the catalogue products derived from those pixels.
      • Those catalogues are incomplete at a measurable rate, and a model trained on both inherits that incompleteness as a systematic.

  • We audit AION-1, a 39-modality transformer trained on more than 200 million objects, using causal interventions on its inputs.

    • EN Key points:
      • arXiv:2608.23626v1 Announce Type: new
      • Abstract: Foundation models for astronomy are trained on survey pixels together with the catalogue products derived from those pixels
      • Those catalogues are incomplete at a measurable rate, and a model trained on both inherits that incompleteness as a systematic
      • We audit AION-1, a 39-modality transformer trained on more than 200 million objects, using causal interventions on its inputs

  • TRACE: Transition-Aware Residual Control for Multi-Objective Materials Discovery

    • Publication Time: 2026-08-27 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2608.23631v1 Announce Type: new.
      • Abstract: Multi-objective materials discovery with LLM agents is often limited not only by how many candidates can be proposed, but by how effectively each costly property evaluation informs the next search step.
      • Existing agents mainly store evaluated candidates and their scores, so they know which materials succeeded but not which executable edits caused useful property changes.
      • This makes local refinement difficult when objectives compete and an edit that improves one property may damage another.
    • EN Key points:
      • arXiv:2608.23631v1 Announce Type: new
      • Abstract: Multi-objective materials discovery with LLM agents is often limited not only by how many candidates can be proposed, but by how effectively each cost…
      • Existing agents mainly store evaluated candidates and their scores, so they know which materials succeeded but not which executable edits caused useful property…
      • This makes local refinement difficult when objectives compete and an edit that improves one property may damage another

  • Function-Level Execution Feedback for Code Preference Optimization

    • Publication Time: 2026-08-27 12:00 Beijing Time

  • Summary: arXiv:2608.23632v1 Announce Type: new.

    • Abstract: Process supervision has improved mathematical reasoning, where intermediate steps are naturally expressed as chains of thought.
    • In code generation, however, process supervision remains underexplored because there is no standard notion of a step.
    • Supervision can target lines, reasoning traces, or program states, making it unclear what to label and optimize.
    • Key Points:
      • arXiv:2608.23632v1 Announce Type: new
      • Abstract: Process supervision has improved mathematical reasoning, where intermediate steps are naturally expressed as chains of thought
      • In code generation, however, process supervision remains underexplored because there is no standard notion of a step
      • Supervision can target lines, reasoning traces, or program states, making it unclear what to label and optimize

  • Auditing the Synthetic Memoir: Measuring Scene-Level Confabulation in LLM-Generated Autobiography Against the Documented Record of the Life It Describes

    • Publication Time: 2026-08-27 12:00 Beijing Time
    • Summary: arXiv:2608.23640v1 Announce Type: new.
      • Abstract: When a large language model (LLM) is asked to write a person’s life, how much of what it writes actually happened?
      • We present a scene-level case-study audit - the first quantified audit of LLM-generated autobiography against a subject-specific ground-truth corpus that we are aware of, based on an unsystematic literature search.
      • The subject and the author of this paper are the same person: a 366-day “page-a-day” book of first-person anecdotal entries was drafted with a conversational LLM whose documented inputs were a template, two exemplar days, and each day’s quote - not her corpus - and every day was subsequently audited at the anecdote-scene level against an independent verification corpus using a four-level rubric fixed before analysis.
    • Key Points:
      • arXiv:2608.23640v1 Announce Type: new

  • Abstract: When a large language model (LLM) is asked to write a person’s life, how much of what it writes actually happened

  • We present a scene-level case-study audit - the first quantified audit of LLM-generated autobiography against a subject-specific ground-truth corpus that we are…

  • The subject and the author of this paper are the same person: a 366-day “page-a-day” book of first-person anecdotal entries was drafted with a conversational LL…

  • How much of a measured AI preference is the model, and how much is the instrument?

    • Publishing Time: 2026-08-27 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2608.23641v1 Announce Type: new.
      • Abstract: Model welfare research infers what a model prefers from the answers returned to prompts written to elicit preferences.
      • (2025), Tagliabue and Dung (2025) and Trhlik et al.
      • (2026) have built four instruments for that purpose, and their findings disagree.
    • EN Key Points:
      • arXiv:2608.23641v1 Announce Type: new
      • Abstract: Model welfare research infers what a model prefers from the answers returned to prompts written to elicit preferences
      • Keeling et al
      • (2024), Mazeika et al

  • AI Agents Push Humans Out of the Loop

    • Publishing Time: 2026-08-27 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2608.23642v1 Announce Type: new.
      • Abstract: AI agents pose significant risks as they are granted increasing autonomy.
      • A commonly proposed solution is human oversight and keeping a ‘‘human in the loop’’, but this is not a simple solution: Not only do current approaches to AI agent design impede effective human oversight, but the cognitive capacities required for it are also themselves degraded by extended use of AI systems.
      • This position paper argues that current approaches to the development and deployment of AI agent systems do not support effective human oversight – they contribute to its degradation.
    • EN Key Points:
      • arXiv:2608.23642v1 Announce Type: new

  • Abstract: AI agents pose significant risks as they are granted increasing autonomy

  • A commonly proposed solution is human oversight and keeping a ‘‘human in the loop’’, but this is not a simple solution: Not only do current approaches to AI age…

  • This position paper argues that current approaches to the development and deployment of AI agent systems do not support effective human oversight – they contri…

  • FLARE: A Systematic, Uncertainty-Aware Framework for Evidence-Based Adoption of Artificial Intelligence in Healthcare

    • Published: 2026-08-27 12:00 Beijing Time
    • Abstract: [Translation Pending] - arXiv:2608.23643v1 Announce Type: new.
      • Abstract: Artificial intelligence is increasingly being introduced into healthcare workflows, yet most evaluations emphasize model accuracy rather than whether adoption is economically worthwhile in real clinical settings.
      • This study proposes FLARE, a systematic and uncertainty-aware framework for evaluating the financial and operational implications of adopting AI in healthcare.
      • FLARE combines fuzzy logic, time-driven activity-based costing, and return on investment analysis to estimate the cost of clinical service delivery, the cost of AI development and operation, and the economic consequences of workflow integration under uncertainty.
    • EN Key Points:
      • arXiv:2608.23643v1 Announce Type: new
      • Abstract: Artificial intelligence is increasingly being introduced into healthcare workflows, yet most evaluations emphasize model accuracy rather than whether…
      • This study proposes FLARE, a systematic and uncertainty-aware framework for evaluating the financial and operational implications of adopting AI in healthcare
      • FLARE combines fuzzy logic, time-driven activity-based costing, and return on investment analysis to estimate the cost of clinical service delivery, the cost of…

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

  • Detection != Reliable Control: Decodable Empathy Directions Yield at Most Partial Shifts in Automated Empathy Scores

    • Publication Time: 2026-08-27 12:00 Beijing Time
    • Abstract: arXiv:2608.24901v1 Announce Type: new.
      • Abstract: A decodable “empathy” direction is routinely read as a causal lever, conflating decodability, automated-metric control, and human-perceived change.
      • We test this for two EPITOME-derived facets – Recognition (cognitive) and Resonance (affective) – in three instruction-tuned LLMs, scoring every intervention with two LLM judges and a discriminative EPITOME classifier, each gated by an emotional-vs-neutral positive control.
      • The control passes for the affective facet across all automated instruments, but cognitive range is inconsistent across them.
    • Key English Points:
      • arXiv:2608.24901v1 Announce Type: new
      • Abstract: A decodable “empathy” direction is routinely read as a causal lever, conflating decodability, automated-metric control, and human-perceived change
      • We test this for two EPITOME-derived facets – Recognition (cognitive) and Resonance (affective) – in three instruction-tuned LLMs, scoring every intervention…
      • The control passes for the affective facet across all automated instruments, but cognitive range is inconsistent across them

  • Semantic Variability of Replies Across LLMs: Implications for Designing Conversation-Based Assessment

    • Publication Time: 2026-08-27 12:00 Beijing Time
    • Abstract: arXiv:2608.24920v1 Announce Type: new.
      • Abstract: This study examines whether LLM-generated replies remain semantically consistent when the underlying LLM changes.
      • Using messages from real collaborative conversations, we compared the semantic similarity of generated replies across LLMs under two conditions: with and without preceding chat history.
      • Results show that model choice and conversational context both affect response similarity and alignment with human replies.
    • Key English Points:

  • arXiv:2608.24920v1 Announce Type: new

  • Abstract: This study examines whether LLM-generated replies remain semantically consistent when the underlying LLM changes

  • Using messages from real collaborative conversations, we compared the semantic similarity of generated replies across LLMs under two conditions: with and withou…

  • Results show that model choice and conversational context both affect response similarity and alignment with human replies

  • The Dialect Tax: Dialectal Biases Persist throughout the Language Modeling Pipeline

    • Published: 2026-08-27 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2608.24952v1 Announce Type: new.
      • Abstract: Systematic dialectal performance gaps in language models (LMs) are well documented, but the source of these disparities within the modern language modeling pipeline remains unclear.
      • Our study traces this “dialect tax” across the natural language processing pipeline.
      • Using parallel English dialect corpora that hold meaning fixed while varying surface form, we first confirm that LMs recognize matched Standard American English (SAE) and dialectal texts as semantically equivalent.
    • EN Highlights:
      • arXiv:2608.24952v1 Announce Type: new
      • Abstract: Systematic dialectal performance gaps in language models (LMs) are well documented, but the source of these disparities within the modern language mod…
      • Our study traces this “dialect tax” across the natural language processing pipeline
      • Using parallel English dialect corpora that hold meaning fixed while varying surface form, we first confirm that LMs recognize matched Standard American English…

  • Unsupervised Post-Training of Foundation Models: A Survey

    • Published: 2026-08-27 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2608.24982v1 Announce Type: new.
      • Abstract: Foundation-model post-training usually relies on human labels, preference data, stronger teachers, or executable verifiers.

  • We study Unsupervised Post-Training (UPT): update-bearing adaptation on unlabeled inputs whose learning signal is derived from same-lineage model artifacts rather than an external oracle.

    • We catalog 80 strict UPT methods and organize them by the object that supplies the update signal: a prediction statistic, a sample relation, a self-generated target, or an internal evaluator.
    • EN Key Points:
      • arXiv:2608.24982v1 Announce Type: new
      • Abstract: Foundation-model post-training usually relies on human labels, preference data, stronger teachers, or executable verifiers
      • We study Unsupervised Post-Training (UPT): update-bearing adaptation on unlabeled inputs whose learning signal is derived from same-lineage model artifacts rath…
      • We catalog 80 strict UPT methods and organize them by the object that supplies the update signal: a prediction statistic, a sample relation, a self-generated ta…

  • Does Fine-Tuning Undo Activation Steering? Behavioural Recovery Without Weight-Edit Reversal

    • Publication Time: 2026-08-27 12:00 Beijing Time
    • Abstract: [Translation Pending] - arXiv:2608.24988v1 Announce Type: new.
      • Abstract: Activation steering can be embedded directly into a language model’s weights, shaping behaviour without inference-time intervention and offering a way to encode alignment prior to release.
      • However, models are routinely fine-tuned after deployment, and it is unknown whether embedded interventions survive this.
      • We study the stability of embedded steering for refusal suppression and brevity induction across five instruction-tuned models (3B-14B) under non-adversarial SFT and RLHF.
    • EN Key Points:
      • arXiv:2608.24988v1 Announce Type: new
      • Abstract: Activation steering can be embedded directly into a language model’s weights, shaping behaviour without inference-time intervention and offering a way…
      • However, models are routinely fine-tuned after deployment, and it is unknown whether embedded interventions survive this

  • We study the stability of embedded steering for refusal suppression and brevity induction across five instruction-tuned models (3B-14B) under non-adversarial SF…

  • The Imperfective Paradox Is Not Necessarily in Large Language Models: A Benchmark Failure Before a Model Failure

    • Publication Time: 2026-08-27 12:00 Beijing Time
    • Abstract: [Translation Pending] - arXiv:2608.25005v1 Announce Type: new.
      • Abstract: The imperfective paradox provides a useful test of compositional semantic analysis.
      • Recent work constructs an NLI benchmark and reports that models frequently infer completed telic events from progressive descriptions, attributing this behavior to a Teleological Bias.
      • It further argues that prompting interventions cause a Calibration Crisis.
    • EN Key Points:
      • arXiv:2608.25005v1 Announce Type: new
      • Abstract: The imperfective paradox provides a useful test of compositional semantic analysis
      • Recent work constructs an NLI benchmark and reports that models frequently infer completed telic events from progressive descriptions, attributing this behavior…
      • It further argues that prompting interventions cause a Calibration Crisis

  • A Primer on Computational Semantics for Artificial Intelligence Systems

    • Publication Time: 2026-08-27 12:00 Beijing Time
    • Abstract: [Translation Pending] - arXiv:2608.25022v1 Announce Type: new.
      • Abstract: As people adopt transformer-based language models (e.g., ChatGPT and Gemini) for an increasing number of use-cases, it is important to know how such models learn and represent the meaning of the language, and to be more informed about what language is.
      • This document is an attempt to help the reader understand how linguistic meaning (i.e., semantics) is approached from different fields of scientific and philosophical examination.

  • I also explain three primary semantic theories: formal semantics, grounded semantics, and distributional semantics then compare how transformer-based language models differ from how humans learn language.

    • EN Highlights:
      • arXiv:2608.25022v1 Announce Type: new
      • Abstract: As people adopt transformer-based language models (e.g., ChatGPT and Gemini) for an increasing number of use-cases, it is important to know how such m…
      • This document is an attempt to help the reader understand how linguistic meaning (i.e., semantics) is approached from different fields of scientific and philoso…
      • I also explain three primary semantic theories: formal semantics, grounded semantics, and distributional semantics then compare how transformer-based language m…

  • Behind the [MASK]: Disentangling Representation and Faithfulness in DAPF-Based Dementia Detection

    • Published: 2026-08-27 12:00 Beijing Time
    • Abstract: [Translation Pending] - arXiv:2608.25028v1 Announce Type: new.
      • Abstract: Spoken-language analysis via prompt-based domain-adaptive models is a promising direction for low-resource, non-invasive dementia screening, but such models remain internally opaque.
      • We study the interpretability of the Domain-Adapted models via Prompt-based Fine-tuning (DAPF) framework, which casts dementia detection as diagnosis-related masked-token prediction.
      • We interpret DAPF and strong baselines using a variety of probing and analysis techniques, finding that DAPF achieved the best overall performance (accuracy=0.83 and macro-F1=0.83) with diagnosis most recoverable from its [MASK] representation.
    • EN Highlights:
      • arXiv:2608.25028v1 Announce Type: new
      • Abstract: Spoken-language analysis via prompt-based domain-adaptive models is a promising direction for low-resource, non-invasive dementia screening, but such…

  • We study the interpretability of the Domain-Adapted models via Prompt-based Fine-tuning (DAPF) framework, which casts dementia detection as diagnosis-related ma…

  • We interpret DAPF and strong baselines using a variety of probing and analysis techniques, finding that DAPF achieved the best overall performance (accuracy=0.8…

  • Padamitra: Grounded Glossary Generation for Classical Sanskrit

    • Published: 2026-08-27 12:00 Beijing Time
    • Summary: [Awaiting Translation] - arXiv:2608.25038v1 Announce Type: new.
      • Abstract: We introduce grounded glossary generation, a structured task requiring models to recover semantically meaningful Sanskrit phrases and produce translation-grounded meanings from a sloka-translation pair, formalizing the traditional patha commentary practice as an evaluable NLP objective.
      • We construct a benchmark of 31,316 sloka-translation-glossary triples from the Valmiki Ramayana and Srimad Bhagavatam, paired with two metrics: Jaccard for phrase recovery and Meaning Faithfulness for semantic consistency.
      • Across zero-shot, few-shot, and instruction fine-tuned variants of Gemma-3n-E4B, Gemma-3-12B, Phi-4, and Qwen3.5-9B, instruction fine-tuning substantially outperforms prompting, while explicit segmentation yields gains.
    • EN Highlights:
      • arXiv:2608.25038v1 Announce Type: new
      • Abstract: We introduce grounded glossary generation, a structured task requiring models to recover semantically meaningful Sanskrit phrases and produce translat…
      • We construct a benchmark of 31,316 sloka-translation-glossary triples from the Valmiki Ramayana and Srimad Bhagavatam, paired with two metrics: Jaccard for phra…
      • Across zero-shot, few-shot, and instruction fine-tuned variants of Gemma-3n-E4B, Gemma-3-12B, Phi-4, and Qwen3.5-9B, instruction fine-tuning substantially outpe…

  • DataKernelBench: Can LLMs Optimize Database Queries on GPUs?

    • Published: 2026-08-27 12:00 Beijing Time
    • Summary: [Awaiting Translation] - arXiv:2608.25061v1 Announce Type: new.

  • Abstract: GPUs increasingly accelerate database systems, but query-specific peak performance still often relies on hand-written kernels.

  • Existing LLM kernel benchmarks focus on machine learning operators, leaving irregular, heterogeneous, data-movement-heavy database-style operators untested.

  • We introduce DataKernelBench, which translates SQL into validated PyTorch TorchPlan programs and evaluates LLMs that optimize either the core tensor-bounded snippet or the full query in CUDA or Triton through execution-guided repair.

  • EN Key Points:

    • arXiv:2608.25061v1 Announce Type: new
    • Abstract: GPUs increasingly accelerate database systems, but query-specific peak performance still often relies on hand-written kernels
    • Existing LLM kernel benchmarks focus on machine learning operators, leaving irregular, heterogeneous, data-movement-heavy database-style operators untested
    • We introduce DataKernelBench, which translates SQL into validated PyTorch TorchPlan programs and evaluates LLMs that optimize either the core tensor-bounded sni…

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

  • Dynamic Influence-Weighted Distillation for Single-IMU Activity Recognition

    • Published: 2026-08-27 12:00 Beijing Time
    • Abstract: [Translation Pending] - arXiv:2608.24904v1 Announce Type: new.
      • Abstract: Inertial sensors at multiple body locations can improve activity recognition, but requiring every sensor at inference increases the deployment burden.
      • We study whether four synchronized IMUs available during training can improve a student that uses only the right-arm IMU during fitting and inference.
      • A frozen four-IMU teacher provides logit and feature targets.
    • EN Key Points:
      • arXiv:2608.24904v1 Announce Type: new
      • Abstract: Inertial sensors at multiple body locations can improve activity recognition, but requiring every sensor at inference increases the deployment burden

  • We study whether four synchronized IMUs available during training can improve a student that uses only the right-arm IMU during fitting and inference

  • A frozen four-IMU teacher provides logit and feature targets

  • GreenLeaf Law Embed Tiny: A Compact Embedding Model for Legal Domain Retrieval

    • Release Time: 2026-08-27 12:00 Beijing Time
    • Abstract: [Pending Translation] - arXiv:2608.24936v1 Announce Type: new.
      • Abstract: We present GreenLeaf Law Embed Tiny, a 0.6B parameter embedding model for legal domain retrieval.
      • GreenLeaf-Tiny achieves 75.11% on the Massive Legal Embedding Benchmark (MLEB) and 64.38% on MTEB(Law, v1),demonstrating competitive performance among models under 1B parameters.
      • Our approach combines a two-stage training pipeline that first distills knowledge from a larger teacher model into a compact student architecture, then applies domain-specific fine-tuning with hard negative mining; a carefully curated dataset of 3.4 million query-passage pairs, including 150,000 human-curated samples across diverse legal jurisdictions; and an efficient inference architecture supporting multiple quantization levels (BF16, INT8, binary) enabling deployment in resource-constrained environments.
    • EN Highlights:
      • arXiv:2608.24936v1 Announce Type: new
      • Abstract: We present GreenLeaf Law Embed Tiny, a 0.6B parameter embedding model for legal domain retrieval
      • GreenLeaf-Tiny achieves 75.11% on the Massive Legal Embedding Benchmark (MLEB) and 64.38% on MTEB(Law, v1),demonstrating competitive performance among models un…
      • Our approach combines a two-stage training pipeline that first distills knowledge from a larger teacher model into a compact student architecture, then applies…

  • Multi-Modal Anomaly Detection: A Survey

    • Release Time: 2026-08-27 12:00 Beijing Time
    • Abstract: [Pending Translation] - arXiv:2608.24937v1 Announce Type: new.

  • Abstract: Multi-Modal Anomaly Detection (MMAD) detects rare abnormal events from heterogeneous data sources and is increasingly used in safety- and reliability-critical applications such as industrial inspection and cybersecurity.

  • Yet the literature is fragmented across domains and modality combinations, and existing surveys usually group methods by architecture rather than by how abnormality is defined and separated in multi-modal settings.

  • We survey MMAD from an assumption-driven perspective.

    • EN Key Points:
      • arXiv:2608.24937v1 Announce Type: new
      • Abstract: Multi-Modal Anomaly Detection (MMAD) detects rare abnormal events from heterogeneous data sources and is increasingly used in safety- and reliability-…
      • Yet the literature is fragmented across domains and modality combinations, and existing surveys usually group methods by architecture rather than by how abnorma…
      • We survey MMAD from an assumption-driven perspective

  • ExFold: Unified Expert Folding for Training-Free MoE Prefill-Decode Acceleration

    • Publication Time: 2026-08-27 12:00 Beijing Time
    • Abstract: [Translation Pending] - arXiv:2608.24938v1 Announce Type: new.
      • Abstract: Mixture-of-Experts (MoE) models scale capacity for strong quality while keeping per-token compute bounded through sparse expert activation.
      • Yet low-latency MoE serving is increasingly challenging, because it spans two inference phases with fundamentally different bottlenecks: prefill is dominated by token-wise expert computation, whereas decode is constrained by memory traffic from the batch-wise activated expert set.
      • However, existing training-free acceleration methods optimize only a single resource proxy, either the experts each token executes or the experts a batch activates, and either discard the excluded experts’ contribution or leave it only implicitly approximated.
    • EN Key Points:
      • arXiv:2608.24938v1 Announce Type: new

  • Abstract: Mixture-of-Experts (MoE) models scale capacity for strong quality while keeping per-token compute bounded through sparse expert activation

  • Yet low-latency MoE serving is increasingly challenging, because it spans two inference phases with fundamentally different bottlenecks: prefill is dominated by…

  • However, existing training-free acceleration methods optimize only a single resource proxy, either the experts each token executes or the experts a batch activa…

  • When Does Frequency Decomposition Benefit Physics-Informed Neural Networks? A Preliminary Ablation Study

    • Publication Time: 2026-08-27 12:00 Beijing Time
    • Abstract: arXiv:2608.24940v1 Announce Type: new.
      • Abstract: Partial differential equations (PDEs) often have high-frequency and multi-scale features that neural networks struggle to approximate.
      • Physics-Informed Neural Networks (PINNs) build the governing equations directly into training, but suffer from spectral bias: they learn low-frequency components faster than high-frequency ones.
      • Techniques such as Fourier feature embeddings and sinusoidal activations address this, but most studies assume they help across the board without checking which spectral regimes actually benefit.
    • Key Points (EN):
      • arXiv:2608.24940v1 Announce Type: new
      • Abstract: Partial differential equations (PDEs) often have high-frequency and multi-scale features that neural networks struggle to approximate
      • Physics-Informed Neural Networks (PINNs) build the governing equations directly into training, but suffer from spectral bias: they learn low-frequency component…
      • Techniques such as Fourier feature embeddings and sinusoidal activations address this, but most studies assume they help across the board without checking which…

  • FAMPWQ: Fisher Information-based Adaptive Mixed Precision Weight Quantization for Effective LLM Inference

    • Publication Time: 2026-08-27 12:00 Beijing Time

  • Abstract: [To be translated] - arXiv:2608.24945v1 Announce Type: new.

    • Abstract: Recent years have witnessed remarkable achievements of Large Language Models (LLMs) in multiple domains, while the excessive resource requirements of LLMs hinder the deployment on resource-constrained devices.
    • Although model quantization stands out as an effective approach, conventional quantization approaches typically incur severe performance degradation due to uniform bit-width or simple heuristic sensitivity evaluation.
    • In this paper, we propose a novel Fisher information-based Adaptive Mixed Precision Weight Quantization approach, i.e., FAMPWQ, which performs layer-adaptive weight quantization for effective LLM inference on commodity GPUs.
    • EN Highlights:
      • arXiv:2608.24945v1 Announce Type: new
      • Abstract: Recent years have witnessed remarkable achievements of Large Language Models (LLMs) in multiple domains, while the excessive resource requirements of…
      • Although model quantization stands out as an effective approach, conventional quantization approaches typically incur severe performance degradation due to unif…
      • In this paper, we propose a novel Fisher information-based Adaptive Mixed Precision Weight Quantization approach, i.e., FAMPWQ, which performs layer-adaptive we…

  • MacroAgent: Regularity-Aware Macro Legalization with LLM-Agent-Designed Contour Algorithms

    • Publication Time: 2026-08-27 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2608.24946v1 Announce Type: new.
      • Abstract: Macros constitute a large part of the core area in modern very large-scale integration (VLSI) designs.
      • Moreover, macro positions have a significant impact on the final quality of result (QoR), and macro legalization is typically the final step in determining the macro positions.
      • However, existing approaches related to macro legalization either lack robustness or incur substantial computational costs or neglect the regularity between macros.
    • EN Highlights:

  • arXiv:2608.24946v1 Announce Type: new

    • Abstract: Macros constitute a large part of the core area in modern very large-scale integration (VLSI) designs
    • Moreover, macro positions have a significant impact on the final quality of result (QoR), and macro legalization is typically the final step in determining the…
    • However, existing approaches related to macro legalization either lack robustness or incur substantial computational costs or neglect the regularity between mac…

  • CAT-GS: Balanced Multimodal Learning via Calibrated Gating and Fusion Surgery

    • Publication Time: 2026-08-27 12:00 Beijing Time
    • Abstract: [To be translated] - arXiv:2608.24947v1 Announce Type: new.
      • Abstract: End-to-end training of multimodal neural networks often exhibits unstable neural dynamics characterized by three coupled failure modes that degrade learning: (i) modality imbalance, where one branch dominates gradient-based optimization; (ii) unstable gating, where noisy confidence cues induce erratic modality selection; and (iii) fusion interference, where modality-specific gradients conflict at the shared fusion layer.
      • We propose CAT-GS (Calibrated, Adaptive, Thresholded Gating with Fusion Surgery), a neural dynamics-based optimization controller for intelligent computing applications.
      • CAT-GS operates during backpropagation without modifying model architectures, fusion modules, or task losses.
    • EN Key Points:
      • arXiv:2608.24947v1 Announce Type: new
      • Abstract: End-to-end training of multimodal neural networks often exhibits unstable neural dynamics characterized by three coupled failure modes that degrade le…
      • We propose CAT-GS (Calibrated, Adaptive, Thresholded Gating with Fusion Surgery), a neural dynamics-based optimization controller for intelligent computing appl…
      • CAT-GS operates during backpropagation without modifying model architectures, fusion modules, or task losses

  • Demystifying Reinforcement Learning Post-Training of Language Models

    • Published: 2026-08-27 12:00 Beijing Time
    • Summary: [Translation pending] - arXiv:2608.24949v1 Announce Type: new.
      • Abstract: Reinforcement learning (RL) post-training has emerged as a powerful framework for enhancing the capabilities of large language models (LLMs), enabling impressive reasoning, math, and coding capabilities.
      • Yet for many researchers and practitioners, the principles behind classical RL remain a “black box”.
      • In this work, we deconstruct the RL post-training algorithm, investigating each step to clarify what is actually happening beneath the surface.
    • EN Key Points:
      • arXiv:2608.24949v1 Announce Type: new
      • Abstract: Reinforcement learning (RL) post-training has emerged as a powerful framework for enhancing the capabilities of large language models (LLMs), enabling…
      • Yet for many researchers and practitioners, the principles behind classical RL remain a “black box”
      • In this work, we deconstruct the RL post-training algorithm, investigating each step to clarify what is actually happening beneath the surface

  • AFDBench: A Reasoning-First AI Scientist for NationalWeather Service Forecast Discussions

    • Published: 2026-08-27 12:00 Beijing Time
    • Summary: [Translation pending] - arXiv:2608.24954v1 Announce Type: new.
      • Abstract: Large language models (LLMs) hallucinate numerical values when generating high-stakes meteorological text, posing risks for weather communication.
      • We present AFDBench, an AI meteorologist that generates professional Area Forecast Discussions (AFDs) by reasoning through structured AI weather forecast data from Google’s WeatherNext 2.

  • We introduce AFDBench, the first benchmark for evaluating generative meteorological reasoning, comprising 7,732 expert written discussions from 13 National Weather Service (NWS) offices paired with real AI weather forecast inputs, and three complementary metrics: Met-Align (numerical accuracy), Style-Align (professional dialect adherence), and Input-Grounding (fidelity to source weather data).

    • EN Key Points:
      • arXiv:2608.24954v1 Announce Type: new
      • Abstract: Large language models (LLMs) hallucinate numerical values when generating high-stakes meteorological text, posing risks for weather communication
      • We present AFDBench, an AI meteorologist that generates professional Area Forecast Discussions (AFDs) by reasoning through structured AI weather forecast data f…
      • We introduce AFDBench, the first benchmark for evaluating generative meteorological reasoning, comprising 7,732 expert written discussions from 13 National Weat…