-
1. The Tasteful Agent: Measuring and Improving Taste in Long-Horizon Tasks
Hugging Face Daily Papers
19 小时前 · 2026/09/23 08:00
111 HF upvotes · 4 comments · 1 GitHub stars
来自 Hugging Face Daily Papers，主题偏「Agent、Reasoning、Eval/Data」。摘要显示它主要讨论 LLM agents increasingly work on long-horizon tasks, and the decisions they make along the way, such as which hypothesis to test or which implementation to build on, determine the outcome of the whole run. Making these decisions well is becoming a key capabilit... 先把它当作时效信号看：判断它是否正在影响 agent、RAG、多模态、post-training、评测或 AI infra 的产品/研究方向。
为什么值得看适合观察 agentic RL、工具调用、工作流自动化或软件代理能力是否出现新方法。
读原文判断如果你要找可复现 demo、开源工具或产品化线索，建议点开项目/GitHub；否则先看中文摘要即可。
AgentReasoningEval/Data
Original abstract
LLM agents increasingly work on long-horizon tasks, and the decisions they make along the way, such as which hypothesis to test or which implementation to build on, determine the outcome of the whole run. Making these decisions well is becoming a key capability for both engineering and research agents. We refer to the ability to make good long-horizon decisions as the taste of an agent. While existing benchmarks measure the end-to-end success of agents on long-horizon tasks, none of them measures the taste of an agent. To address this problem, we build Taste-Bench, a benchmark of taste questions constructed automatically from trajectories that agents produced in engineering and research tasks. Each question presents a decision fork, a point in a trajectory where multiple directions are available and one of them leads to a better outcome, and the evaluated model chooses among these directions without seeing what happens after the fork. We mine these forks automatically from parallel attempts at the same task and from detours inside a single trajectory, without needing human annotation. We evaluate frontier models on Taste-Bench and find that the best model answers only 59.7% of the questions correctly. We further find that forks whose deciding evidence appears later in the trajectory are much harder for every model, and that a larger reasoning budget does not improve the accurac...
-
2. RULER: Instance-aware Rubric Rewards for SVG Generation
Hugging Face Daily Papers
19 小时前 · 2026/09/23 08:00
59 HF upvotes · 2 comments
来自 Hugging Face Daily Papers，主题偏「Multimodal、Post-training/Alignment、Eval/Data」。摘要显示它主要讨论 Generating Scalable Vector Graphics (SVG) code from natural-language instructions is an open-ended task without absolute visual ground truth, leaving both evaluation and policy optimization without a faithful signal. Scalar metrics (CLIP, Aesthetic) calibrated... 先把它当作时效信号看：判断它是否正在影响 agent、RAG、多模态、post-training、评测或 AI infra 的产品/研究方向。
为什么值得看适合观察图像、视频、语音和科学多模态任务是否出现新的产品能力边界。
读原文判断如果你要找可复现 demo、开源工具或产品化线索，建议点开项目/GitHub；否则先看中文摘要即可。
MultimodalPost-training/AlignmentEval/DataCode
Original abstract
Generating Scalable Vector Graphics (SVG) code from natural-language instructions is an open-ended task without absolute visual ground truth, leaving both evaluation and policy optimization without a faithful signal. Scalar metrics (CLIP, Aesthetic) calibrated on natural images transfer poorly to stylized vector content, and reusing them as RL rewards triggers reward hacking. We address both limitations with rubric-based scoring. We first establish empirically that prompting a vision-language judge with a multi-axis rubric correlates with human judgments far better than scalar metrics, both across samples and within instructions. Building on this finding, we introduce RULER (Instance-aware Rubric Rewards for Reinforcement Learning), which converts each instruction into an instance-aware rubric of six items spanning semantic, visual, and stylistic axes; a judge VLM scores rendered rollouts item-by-item, and the weighted satisfactions form a fine-grained reward optimized via Group Relative Policy Optimization. Because the rubric is derived from text alone, RULER requires neither paired SVG ground truth nor human preference labels. On MMSVG-Illustration and MMSVG-Icon, RULER lifts the rubric score from 0.432/0.395 to 0.693/0.683, surpassing dedicated SVG specialists and matching the substantially larger DeepSeek-V3, with ablations identifying rubric design as the active lever for...
-
3. All-in-One Multilingual Scene Text Recognition with Script-aware Mixture-of-Experts
Hugging Face Daily Papers
19 小时前 · 2026/09/23 08:00
35 HF upvotes · 2 comments · 6 GitHub stars
来自 Hugging Face Daily Papers，主题偏「Multimodal、Eval/Data、AI Infra」。摘要显示它主要讨论 Multilingual scene text recognition (STR) remains challenging due to the scarcity of training data for most languages and the difficulty of serving diverse scripts within a single model. Existing solutions either deploy one recognizer per language, inflating c... 先把它当作时效信号看：判断它是否正在影响 agent、RAG、多模态、post-training、评测或 AI infra 的产品/研究方向。
为什么值得看适合观察图像、视频、语音和科学多模态任务是否出现新的产品能力边界。
读原文判断如果你要找可复现 demo、开源工具或产品化线索，建议点开项目/GitHub；否则先看中文摘要即可。
MultimodalEval/DataAI InfraCode
Original abstract
Multilingual scene text recognition (STR) remains challenging due to the scarcity of training data for most languages and the difficulty of serving diverse scripts within a single model. Existing solutions either deploy one recognizer per language, inflating cost and introducing error accumulation, or rely on massive vision-language models (VLMs) that are expensive and still inaccurate on many scripts. In this work, we pursue an all-in-one multilingual recognizer that is simpler than per-language experts, lighter than VLMs, and more accurate than both. First, we construct TextMuSS-10M, a large-scale synthetic scene text dataset spanning 10 scripts and 229 languages. It provides balanced and sufficient supervision where real data is unavailable. Second, we propose ScriptMoE, a script-aware Mixture-of-Experts (MoE) architecture. It shares a single visual encoder and replaces the dense decoder with a sparse MoE block, which consists of an image-level router dispatches each image to the top-2 script-aligned experts and a shared expert absorbs cross-script knowledge. Extensive experiments on our assembled TextMuSS-Bench (10 scripts, 10,899 images) show that ScriptMoE achieves the highest accuracy of 82.06%, outperforming the strongest STR baseline by 1.31%. On the CC-OCR end-to-end multilingual task, replacing only the recognizer in PP-OCRv5 with ScriptMoE lifts F1 score from 65.71%...
-
4. Ovis-Embedding: Pushing the Frontiers of Universal Omni-Modal Embeddings
Hugging Face Daily Papers
19 小时前 · 2026/09/23 08:00
19 HF upvotes · 23 GitHub stars
来自 Hugging Face Daily Papers，主题偏「RAG/Memory、Multimodal、Eval/Data」。摘要显示它主要讨论 In this report, we introduce Ovis-Embedding, a state-of-the-art omni-modal embedding family built on native integration of text, image, video, and audio. Instead of assembling separate modality towers, Ovis-Embedding uses a shared multimodal backbone to encode... 先把它当作时效信号看：判断它是否正在影响 agent、RAG、多模态、post-training、评测或 AI infra 的产品/研究方向。
为什么值得看适合观察知识工作、企业搜索、长期记忆和本地资料库产品的新实现路径。
读原文判断如果你要找可复现 demo、开源工具或产品化线索，建议点开项目/GitHub；否则先看中文摘要即可。
RAG/MemoryMultimodalEval/DataAI InfraCode
Original abstract
In this report, we introduce Ovis-Embedding, a state-of-the-art omni-modal embedding family built on native integration of text, image, video, and audio. Instead of assembling separate modality towers, Ovis-Embedding uses a shared multimodal backbone to encode different modalities in a common representation space. Specifically, we make three key advances: (1) native omni-modal initialization: we adopt a pretrained Qwen-omni model as the embedding backbone and adapt it through contrastive training with low-rank initialization; (2) data-centric omni-modal training: we construct a broad, high-quality corpus spanning text, images, video, audio, and interleaved multimodal data. To improve data efficiency, we introduce homogeneous-source sampling to form task-consistent batches with informative in-batch negatives; and (3) embedding-specific training and inference optimization: we use focal loss to emphasize hard examples and similarity-based Embedding Distillation to transfer fine-grained similarity structure from complementary experts. At inference time, low-rank feature decomposition enables compact embeddings with flexible dimensionality and minimal performance loss. Empirical evaluations show that the Ovis-Embedding family achieves state-of-the-art performance on MMEB-v3, MMEB-v2, MVEB, MAEB, and RTEB, demonstrating its effectiveness across text, image, video, and audio modalitie...
-
5. From Pattern Recognizers to Personalized Companions: A Survey of Large Language Models in Mental Health
Hugging Face Daily Papers
19 小时前 · 2026/09/23 08:00
15 HF upvotes · 2 comments · 127 GitHub stars
来自 Hugging Face Daily Papers，主题偏「Agent、RAG/Memory、Reasoning」。摘要显示它主要讨论 The rising global prevalence of mental health conditions, together with longstanding barriers in traditional healthcare, such as limited resources, high cost, stigma, and privacy concerns, has created an urgent need for accessible and scalable support. Large L... 先把它当作时效信号看：判断它是否正在影响 agent、RAG、多模态、post-training、评测或 AI infra 的产品/研究方向。
为什么值得看适合观察 agentic RL、工具调用、工作流自动化或软件代理能力是否出现新方法。
读原文判断如果你要找可复现 demo、开源工具或产品化线索，建议点开项目/GitHub；否则先看中文摘要即可。
AgentRAG/MemoryReasoningEval/Data
Original abstract
The rising global prevalence of mental health conditions, together with longstanding barriers in traditional healthcare, such as limited resources, high cost, stigma, and privacy concerns, has created an urgent need for accessible and scalable support. Large Language Models (LLMs) have emerged as a transformative technology with strong potential to democratize mental health support through advanced natural language understanding and generation. However, the rapidly expanding, fragmented body of work in this area lacks a coherent evolutionary narrative, making it difficult to contextualize current progress and identify future directions. This survey addresses this gap by organizing and analyzing the literature around a central thesis: the role of LLMs in mental health is evolving through three distinct, increasingly sophisticated phases. We trace this trajectory from Phase I, in which LLMs act primarily as passive Information Tools and Pattern Recognizers for assessment; through Phase II, where they function as Empathetic Conversationalists for in-the-moment, stateless interactions; to the current frontier, Phase III, which seeks Longitudinal, Personalized Companions implemented as stateful cognitive agents. To support this framework, we systematically review core technologies, agent architectures (Profile, Memory, Reasoning, and Planning), and the critical infrastructure of dat...
-
6. Bellman Policy Optimization
Hugging Face Daily Papers
19 小时前 · 2026/09/23 08:00
24 HF upvotes · 2 comments
来自 Hugging Face Daily Papers，主题偏「Reasoning、Post-training/Alignment、Eval/Data」。摘要显示它主要讨论 Reinforcement learning with verifiable rewards (RLVR) improves the reasoning capabilities of large language models (LLMs). We introduce Bellman Policy Optimization (BPO), a critic-free method derived from Policy Mirror Descent (PMD). For autoregressive generat... 先把它当作时效信号看：判断它是否正在影响 agent、RAG、多模态、post-training、评测或 AI infra 的产品/研究方向。
为什么值得看适合观察模型推理、规划、验证器和复杂任务能力是否有可复用技术路线。
读原文判断先作为日更雷达记录；除非与你当前研究/产品问题高度相关，不必立刻精读全文。
ReasoningPost-training/AlignmentEval/Data
Original abstract
Reinforcement learning with verifiable rewards (RLVR) improves the reasoning capabilities of large language models (LLMs). We introduce Bellman Policy Optimization (BPO), a critic-free method derived from Policy Mirror Descent (PMD). For autoregressive generation with terminal rewards, BPO uses the Bellman equations to reformulate PMD as a trajectory-level objective. The reformulation avoids estimating state values at intermediate states. We prove that it has the same unique optimal solution as the original PMD objective. We derive the practical BPO loss by approximating this objective. Its mismatch-correction weight is a smoothed ratio of complementary token probabilities. Experiments on mathematical reasoning benchmarks demonstrate the effectiveness of BPO.
-
7. Circuit Hypernetworks for Quantum-Augmented Diffusion Language Models
Hugging Face Daily Papers
19 小时前 · 2026/09/23 08:00
22 HF upvotes · 1 comments
来自 Hugging Face Daily Papers，主题偏「RAG/Memory、Eval/Data」。摘要显示它主要讨论 Language models can be adapted by changing the computations applied to individual tokens. Quantum circuits offer one such approach, but evaluating wider circuits inside a large model can be computationally demanding. Here we introduce HyperQ, which adds token-... 先把它当作时效信号看：判断它是否正在影响 agent、RAG、多模态、post-training、评测或 AI infra 的产品/研究方向。
为什么值得看适合观察知识工作、企业搜索、长期记忆和本地资料库产品的新实现路径。
读原文判断如果标题正好贴近当前产品方向，值得点开原文看方法和实验设置；泛读时先存为观察项。
RAG/MemoryEval/Data
Original abstract
Language models can be adapted by changing the computations applied to individual tokens. Quantum circuits offer one such approach, but evaluating wider circuits inside a large model can be computationally demanding. Here we introduce HyperQ, which adds token-conditioned quantum residual branches to a frozen masked-diffusion language model. A quantum residual branch is a module in each transformer block that reads a token's hidden state, emits the coordinates of that token's circuit, executes it, and adds the measured values back through a residual connection. The backbone remains frozen, and only the added branches are trained. Within each branch, a lightweight circuit hypernetwork emits token-specific rotation angles, coupling strengths, and measurement axes in a shared sparse circuit structure. The required expectation values have an exact classical expression whose evaluation cost grows linearly with the qubit count, enabling circuits from 16 to 64 qubits to be trained within a 1.1-billion-parameter backbone. Across downstream benchmarks, increasing circuit width raises the average score from 47.65 to 54.30. At 64 qubits, HyperQ exceeds the backbone and its low-rank-adapted counterpart by 4.71 and 3.67 points, respectively. HyperQ is fine-tuned on 20,000 prompt-response pairs, compared with 200,000 for the classical baselines. These findings support token-conditioned circui...
-
8. Flash-dLLM: IO-Aware KV Caching and Parallel Decoding for Fast, Memory-Efficient Diffusion LLMs
Hugging Face Daily Papers / arXiv Recent AI/ML
1 天前 · 2026/09/23 01:59
13 HF upvotes · 1 comments · 8 GitHub stars
来自 arXiv Recent AI/ML，主题偏「RAG/Memory、Reasoning、Eval/Data」。摘要显示它主要讨论 Diffusion Large Language Models (dLLMs) have recently emerged as a promising alternative to autoregressive LLMs by enabling non-autoregressive text generation. However, their practical deployment remains limited by inefficient inference, largely due to the abs... 先把它当作时效信号看：判断它是否正在影响 agent、RAG、多模态、post-training、评测或 AI infra 的产品/研究方向。
为什么值得看适合观察知识工作、企业搜索、长期记忆和本地资料库产品的新实现路径。
读原文判断如果标题正好贴近当前产品方向，值得点开原文看方法和实验设置；泛读时先存为观察项。
RAG/MemoryReasoningEval/DataAI InfraCode
Original abstract
Diffusion Large Language Models (dLLMs) have recently emerged as a promising alternative to autoregressive LLMs by enabling non-autoregressive text generation. However, their practical deployment remains limited by inefficient inference, largely due to the absence of effective Key-Value (KV) caching and scalable parallel decoding mechanisms. Existing acceleration methods typically study KV caching and parallel decoding in isolation, overlooking the I/O bottlenecks that arise when cache reuse and parallel token verification are jointly applied. In this work, we introduce $\textbf{Flash-dLLM}$, a training-free inference acceleration framework for fast and memory-efficient dLLMs. Flash-dLLM first identifies GPU memory I/O as a dominant bottleneck in KV-cache-enabled dLLM inference and addresses it with an I/O-aware fused KV-cache kernel that reduces redundant memory movement. Building on this optimized cache mechanism, Flash-dLLM further proposes an efficient KV-cache-driven draft-and-verify decoding strategy, where the dLLM itself serves as both drafter and verifier without requiring an auxiliary model. This unified design enables faster decoding while preserving generation quality and improving scalability to longer sequences and larger batch size. Extensive experiments on mathematical reasoning and code-generation benchmarks demonstrate that Flash-dLLM consistently outperforms...
-
9. JEV-as-a-Judge: Accept When Confident, Escalate When Unsure
Hugging Face Daily Papers
19 小时前 · 2026/09/23 08:00
12 HF upvotes · 1 comments
来自 Hugging Face Daily Papers，主题偏「Post-training/Alignment、Eval/Data、AI Infra」。摘要显示它主要讨论 LLM-as-a-judge enables evaluation across diverse tasks, but inference cost and confidence reliability become critical at scale. We study whether a decision-only judge can provide an economical first pass and identify when stronger evaluation is needed. Compari... 先把它当作时效信号看：判断它是否正在影响 agent、RAG、多模态、post-training、评测或 AI infra 的产品/研究方向。
为什么值得看适合观察后训练、RL、偏好优化和安全对齐对模型能力的实际影响。
读原文判断如果你要找可复现 demo、开源工具或产品化线索，建议点开项目/GitHub；否则先看中文摘要即可。
Post-training/AlignmentEval/DataAI Infra
Original abstract
LLM-as-a-judge enables evaluation across diverse tasks, but inference cost and confidence reliability become critical at scale. We study whether a decision-only judge can provide an economical first pass and identify when stronger evaluation is needed. Comparing jev-as-a-judge with sixteen generative and reward-model judges, with blinded human adjudication, we find it within three percentage points of a state-of-the-art LLM judge, our strongest comparator, on ordinary preference and evidence-grounded factuality at 0.36% of the comparator's fee. Larger gaps arise when judgments require checking a derivation or resisting an elaborately written wrong answer. On several benchmarks, JEV's gap to this comparator is concentrated in low-confidence decisions. A frozen cascade that accepts confident verdicts and escalates uncertain ones retains 99% of the comparator's accuracy at lower cost.
-
10. RoboFollow: Unveiling the Instruction Following Mirage in Embodied Agents
Hugging Face Daily Papers
19 小时前 · 2026/09/23 08:00
4 HF upvotes · 1 comments · 2 GitHub stars
来自 Hugging Face Daily Papers，主题偏「Agent、RAG/Memory、Multimodal」。摘要显示它主要讨论 Modern embodied agents achieve impressive success rates, yet their actual instruction-following ability is far weaker than these numbers suggest. We trace this illusion to a structural property we term low scene entropy: when a visual scene admits only one val... 先把它当作时效信号看：判断它是否正在影响 agent、RAG、多模态、post-training、评测或 AI infra 的产品/研究方向。
为什么值得看适合观察 agentic RL、工具调用、工作流自动化或软件代理能力是否出现新方法。
读原文判断如果你要找可复现 demo、开源工具或产品化线索，建议点开项目/GitHub；否则先看中文摘要即可。
AgentRAG/MemoryMultimodalEval/DataCode
Original abstract
Modern embodied agents achieve impressive success rates, yet their actual instruction-following ability is far weaker than these numbers suggest. We trace this illusion to a structural property we term low scene entropy: when a visual scene admits only one valid task, language becomes redundant and a policy can score highly while barely using it. We introduce RoboFollow, a diagnostic benchmark with three principles: (1) High Scene Entropy: each training scene supports multiple kinematically distinct task branches, making vision alone insufficient and forcing reliance on language. (2) Hierarchical Diagnostic Protocol: a four-level protocol (L0--L3) progressively perturbs visual layout and semantics, probing whether equivalent instructions yield consistent behavior and distinct ones yield discriminable behavior across spatial relations, attributes, trajectory constraints, and logic. (3) Confound-Controlled Diagnosis: we simplify interaction objects, restrict actions to the trained repertoire and report stage-wise Intent and Execution scores, isolating comprehension from motor execution. Evaluation of nine VLA and WAM policies shows that strong L0 performance, where attained, does not reliably transfer to L1--L3 under our fine-tuning setup. Representative mitigations, including stronger VLM backbones, QA co-training, LangForce, and Classifier-Free Guidance, all fail to close this...
-
11. Recursive self-improvement of AI research agents
Hugging Face Daily Papers
19 小时前 · 2026/09/23 08:00
6 HF upvotes
来自 Hugging Face Daily Papers，主题偏「Agent、RAG/Memory、Eval/Data」。摘要显示它主要讨论 AI agents are beginning to automate research and development across the AI stack, from improving training efficiency to optimizing inference. A natural next step is to improve the research efficiency of the agents themselves. When an AI research agent's own co... 先把它当作时效信号看：判断它是否正在影响 agent、RAG、多模态、post-training、评测或 AI infra 的产品/研究方向。
为什么值得看适合观察 agentic RL、工具调用、工作流自动化或软件代理能力是否出现新方法。
读原文判断如果标题正好贴近当前产品方向，值得点开原文看方法和实验设置；泛读时先存为观察项。
AgentRAG/MemoryEval/DataAI InfraCode
Original abstract
AI agents are beginning to automate research and development across the AI stack, from improving training efficiency to optimizing inference. A natural next step is to improve the research efficiency of the agents themselves. When an AI research agent's own code is the object of optimization, each accepted rewrite becomes the agent that the next round edits. We refer to this loop as recursive self-improvement. Its significance lies in a long-standing trend, in which increased cumulative spending on R&D yields diminishing returns. Sustained self-improvement offers a way to counter this trend. We present AIDE^2, a system that implements this loop for a frontier AI research agent. It proposes changes to its own code, benchmarks modified versions of itself on a suite of AI R&D tasks, and keeps the changes that perform best on hidden evaluations. In an autonomous 8-day run, AIDE^2 discovered seven successive improvements, ranging from a new search policy to memory mechanisms that compress and manage the agent's growing context. These gains generalize to four held-out benchmarks spanning machine learning engineering, heuristic algorithm engineering, and physics-based weather forecasting, the last of which is out of distribution from the selection tasks. On all four, the strongest discovered agent matches or exceeds a human-engineered production research agent that ranks among the str...
-
12. Tri-PvP: Exposing Modality Bias in Omni-Modal Large Language Models through Perceptual-Propositional Evidence Conflicts
Hugging Face Daily Papers
19 小时前 · 2026/09/23 08:00
2 HF upvotes · 1 comments · 2 GitHub stars
来自 Hugging Face Daily Papers，主题偏「Multimodal、Eval/Data、Code」。摘要显示它主要讨论 Omni-modal large language models (OLLMs) jointly process vision, audio, and text, yet their modality bias under cross-modal conflict remains underexplored. Existing benchmarks conflate two distinct forms of evidence within a single modality: perceptual signals... 先把它当作时效信号看：判断它是否正在影响 agent、RAG、多模态、post-training、评测或 AI infra 的产品/研究方向。
为什么值得看适合观察图像、视频、语音和科学多模态任务是否出现新的产品能力边界。
读原文判断如果你要找可复现 demo、开源工具或产品化线索，建议点开项目/GitHub；否则先看中文摘要即可。
MultimodalEval/DataCode
Original abstract
Omni-modal large language models (OLLMs) jointly process vision, audio, and text, yet their modality bias under cross-modal conflict remains underexplored. Existing benchmarks conflate two distinct forms of evidence within a single modality: perceptual signals (e.g., a photograph or recording of a dog) and propositional signals (e.g., the declarative claim "this is a dog"), such that any measured modality bias is inherently confounded with evidence-form bias, precluding clean attribution to either source. To address this, we introduce Tri-PvP, an 8,000-sample tri-modal conflict benchmark crossing vision, audio, and text, where vision and audio each take perceptual or propositional form. Evaluating five OLLMs, we find robust visual bias across most models and evidence-type conditions. Crucially, we reveal a systematic asymmetry in evidence-form bias: models exhibit a stronger bias toward perceptual signal in vision but propositional in audio. Further analyses via layer-wise linear probing and contrastive decoding reveal that modality bias is already linearly decodable from early representation layers and can only be partially mitigated, calling for mitigation strategies beyond surface-level interventions.
