Search: “verbalized-sampling”

Smart search ranks by meaning as well as keywords (one row per story, last 45 days).

Studying Image Tokenizers as Visual Languages in Unified Multimodal Models

A controlled autoregressive testbed shows validation losses must be analyzed per task, and image tokenizer choice affects joint multimodal text modeling.

Researchers built a pure-autoregressive testbed to study image tokenizers as the 'visual language' of unified multimodal models, tracking task-specific validation losses during multimodal continual pretraining across text, image, text-to-image (T2I), and image-to-text (I2T) prediction. They found that losses exhibit distinct scaling behavior per task and rank tokenizers differently, and that I2T loss over a shared text vocabulary gives a more consistent loss–performance signal than T2I loss. Better reconstruction does not necessarily yield lower task-specific losses or stronger downstream performance, and tokenizer choice can affect text modeling under joint optimization. Case studies examine the discriminator, semantic supervision, and vocabulary size design axes.

Hugging Face daily papers·Sep 7, 20:00 UTC · 8d agoAI research1

40

Studying Image Tokenizers as Visual Languages in Unified Multimodal Models

A controlled pure-autoregressive testbed shows task-specific validation losses rank image tokenizers differently, with I2T loss the most consistent signal.

Researchers built a controlled pure-autoregressive testbed and tracked task-specific validation losses during multimodal continual pretraining across text, image, text-to-image (T2I), and image-to-text (I2T) prediction. They find losses should be analyzed per task because they exhibit distinct scaling behavior and rank tokenizers differently, and that the loss-performance relationship depends on the predicted token space. I2T loss, computed over a shared text vocabulary, correlates consistently with both generation and visual understanding performance after supervised finetuning. Case studies revisit the discriminator, semantic supervision, and vocabulary size as tokenizer design axes.

arXiv cs.AI / cs.LG / cs.CL·Sep 8, 17:57 UTC · 7d agoAI research1

30

ECHO: A Matched-Contrast Benchmark for Context-Sensitive Turn-Taking in Full-Duplex Dialogue

ECHO benchmark pairs identical-overlap Chinese dialogue examples with contrasting contexts, revealing most full-duplex systems bias toward yielding the floor.

ECHO is a paired diagnostic benchmark for Chinese full-duplex turn-taking that matches examples with identical overlap transcripts but contrasting preceding multi-turn contexts, one requiring Yield and the other Keep. It also includes off-talk examples for diagnosing unnecessary yielding and introduces pair accuracy, which grants no credit to constant-action policies. Experiments across multiple full-duplex systems show most exhibit a pronounced Yield bias, performing substantially better on interruptions than backchannels. The benchmark and metadata will be publicly released.

arXiv cs.AI / cs.LG / cs.CL·Sep 15, 15:58 UTC · 20h agoAI research

40

Why Is Video Still So Expensive? A Survey of Inference-Efficiency Mechanisms in Video and Audiovisual LLMs

A survey catalogs inference-efficiency techniques for video and audiovisual LLMs, mapping bottlenecks in sampling, encoding, token reduction, and LLM decoding.

This survey covers inference-efficiency mechanisms for visual and audiovisual video LLMs, reporting reductions in parameters, FLOPs, latency, memory, and token counts. It organizes methods by pipeline stage, covering frame sampling, modality encoding, connector-level token reduction, and LLM prefilling and decoding for systems built since late 2022. The authors compile accuracy-cost comparisons under shared host models and input protocols, identify gaps in audiovisual efficiency and standardized evaluation, and maintain a public repository.

Hugging Face daily papers·Sep 8, 20:00 UTC · 7d agoAI research

40

Bag of Tricks or Bag of Myths? Reducing Modeling Complexity with Task Knowledge in Explainable Suicide Risk Assessment

Researchers audit 31 NLP techniques for clinician-annotated suicide risk prediction, finding only 5 of 31 comparisons yield reliable gains.

A study of 1,635 clinician-annotated social media posts ran roughly 300 controlled experiments across 7 methodological families, auditing techniques such as model scaling, synthetic data, ensembling, and threshold tuning under severe class imbalance. The proposed system reformulates risk factor prediction as entailment between posts and codebook definitions, using architecturally diverse ensembles with class-balanced training and deployment-consistent calibration. It scores 0.8203 for risk, 0.7953 for evidence, and 0.7045 macro-F1 for factors, ranking third among 53 teams.

arXiv cs.AI / cs.LG / cs.CL·Sep 7, 17:14 UTC · 8d agoAI research

14

Quenched Ensemble Sampling

Quenched Ensemble Sampling generalizes nested sampling's hard energy constraint to repulsive potentials, traversing first-order phase transitions where tempering fails.

Quenched Ensemble Sampling generalizes nested sampling's hard energy constraint into a family of repulsive potentials at the energy boundary, preserving monotone energy descent while making the constrained target amenable to scalable gradient-based kernels. On synthetic phase-transition models it estimates marginal likelihood and draws posterior samples across first-order transitions where popular alternatives such as tempering fail. Applications include marginal likelihood estimation for Bayesian neural network architecture comparison and partition function estimation in a high-dimensional continuous lattice field theory.

arXiv cs.AI / cs.LG / cs.CL·Sep 14, 17:16 UTC · 1d agoAI research1

15

Unifying Conformal Language Tasks with In-Context Ensembles

Researchers propose Conformal Relevance, which builds conformal score functions via in-context example curation and ensembling to improve conciseness across seven NLP tasks.

The paper targets NLP tasks like summarization and extractive QA that reduce to retrieving content under coverage and conciseness constraints. Conformal Relevance replaces hand-engineered LLM scoring prompts with curated in-context examples and ensembles, maintaining coverage guarantees while improving conciseness with minimal manual input. The authors demonstrate the framework on seven NLP tasks and contribute theory, including a complementarity condition for when ensembling improves worst-case sentence scores and a saturation bound on ensemble gains.

Hugging Face daily papers·Sep 1, 20:00 UTC · 14d agoAI research1

42

What Else Needs Fixing? Exploring Cost-Effective Test-Time Compute for Revision Propagation in Artifacts Generated Through Conversation

A new benchmark shows LLMs reach 68.3-93% accuracy propagating local revisions across conversationally generated artifacts, with parallel-sample selection most cost-effective.

The paper introduces a benchmark for revision propagation: when users request a local change, LLMs must identify dependencies and update all affected parts of an artifact generated through conversation, where context lives in the chat history. Nine revision methods, including sequential reflection and parallel sampling variants, were evaluated on gpt-oss-20b/120b, gpt-5.4-mini, and qwen3.5-9b/27b/122b. Baselines scored 68.3-93% accuracy, and selecting among three parallel samples via LLM-based or medoid selection improved accuracy by 2.2-9.7% as the most cost-effective test-time compute strategy. Code and dataset are released.

Hugging Face daily papers·Sep 2, 20:00 UTC · 13d agoAI research

30

Verifiable by Construction: Claim-Level Evaluation of Verbatim Citation in Clinical Question Answering

Evaluation of twelve LLMs on 222 clinical questions shows verbatim quotes rarely substantiate claims; claude-opus-5 fully substantiates only 37.1%.

The authors build a standardized harness over four clinical practice guidelines and evaluate twelve LLMs on 222 synthetic clinical questions, measuring citation attachment, verbatim quote production, and claim substantiation. Most models attach verbatim quotes to over 90% of claims from prompting alone, though lightweight models like claude-haiku-4.5 struggle. Quotes frequently fail to substantiate claims: claude-opus-5 quotes 98.0% of claims but fully substantiates only 37.1%, exposing a capability gap for verifiable clinical QA.

arXiv cs.AI / cs.LG / cs.CL·Sep 14, 17:53 UTC · 1d agoAI research

38

From Reweighting to Rewriting: Unlocking the Intervention Effects of Influential Samples in Training Data Attribution

Study shows rewriting responses of influence-selected training examples shifts LLM behavior more strongly than reweighting the same samples.

The paper examines training data attribution, arguing that influence functions identify high-leverage examples whose value goes unrealized under conventional weight-based reweighting interventions. It introduces influence-guided response rewriting, which replaces the responses of influence-selected examples with behavior-aligned or behavior-opposed supervision while keeping instructions fixed, tested across four open-weight LLMs using epistemic abstention as the primary testbed. Rewriting produces stronger, more persistent, and bidirectional behavioral shifts, including on safety refusal, while reweighting the same examples yields weak, inconsistent effects. The results motivate intervention-aware evaluation of TDA methods.

Hugging Face daily papers·Sep 1, 20:00 UTC · 14d agoAI research

30

Molecular Déjà Vu: Digit-Level Retrieval of Published Values in Frontier Language Models

Audit of 22 frontier models finds widespread verbatim retrieval of published molecular property values, with higher reasoning increasing recall of memorized numbers.

An arXiv audit tests 22 frontier LLMs across 12 molecular regression benchmarks for verbatim retrieval of published values. More than 50% of the LLMs show verbatim retrieval on five datasets, and identical experiments are flagged 89% more often at a high reasoning level than at the lowest one. Suppressing retrieval moves model prediction errors closer together in relative terms, suggesting predictive capability is not determined solely by memorized values.

arXiv cs.AI / cs.LG / cs.CL·Sep 4, 17:32 UTC · 11d agoAI research1

42

Inoculation Midtraining with Learned Neologisms

Inoculation Midtraining confines unsafe LLM behavior to a neologism-marked context, reducing misalignment after unsafe post-training but leaking under nearby contextual cues.

The paper introduces Inoculation Midtraining, which teaches a base model during midtraining that unsafe behavior belongs to a context marked by a learned neologism token, then post-trains on unsafe data within that context. Across supervised fine-tuning and RL post-training regimes, the technique reduces misalignment while preserving transfer of benign properties like German or Shakespearean prose. However, it does not outperform standard Inoculation Prompting, is sensitive to training configuration, and produces a leaky boundary that nearby contextual cues can reactivate. The authors conclude it is not yet a load-bearing component of a developer safety framework.

arXiv cs.AI / cs.LG / cs.CL·Sep 14, 17:12 UTC · 1d agoAI safety & security

42

K-Bench: a clinically calibrated benchmark for evaluating large language models in high-risk mental health conversations

Clinician-calibrated K-Bench evaluates 125 LLM configurations on 200 high-risk mental health vignettes, exposing wide variation in suicide and violence risk handling.

K-Bench is a clinician-calibrated, protected benchmark evaluating 125 model configurations from 33 base models across 14 providers on 200 multi-turn vignettes covering suicide, self-harm, domestic violence, substance misuse and no-risk presentations. A frozen GPT-4o judge achieved 94.2% exact agreement with clinician consensus across 6,751 eligible comparisons from 151 clinician-rated transcripts. Leading models combined supportive conversation with combined-risk scores above 95, while risk exploration varied substantially among weaker configurations; therapeutic prompting helped weaker models and elevated reasoning produced no average improvement. A continuously updated public leaderboard is hosted at k-bench.ai with protected test materials.

arXiv cs.AI / cs.LG / cs.CL·Sep 14, 16:49 UTC · 1d agoAI safety & security

46

Studying Without a Syllabus: Task-Agnostic Environment Preprocessing

Paper formalizes task-agnostic environment preprocessing, where agents study unfamiliar environments under a budget to build reusable artifacts for a frozen solver.

The paper formalizes task-agnostic environment preprocessing, where a studying system explores an environment under a budget and produces artifacts like indices, scripts, or procedural guidance for a frozen solver, without task examples or evaluation feedback. The authors compare unaided and archive-equipped meta-agents against fixed synthetic-practice and corpus-processing methods across six heterogeneous benchmarks. A meta-agent variant achieves the highest Avg@3 reward on five benchmarks, while fixed corpus processing remains best on the largest corpus benchmark. Studied artifacts reduce the test-time sampling needed to reach a given score, shifting computation from repeated test-time attempts to a pre-task study phase.

Hugging Face daily papers·Sep 8, 20:00 UTC · 7d agoAI research

20

Technical Manual for a Toolkit for Measuring Contextual Individuation in Transformer Language Models

An open methodology toolkit measures whether transformer language models contextualize fixed word forms across domains using bridge forms and layer-wise silhouette analysis.

The manual documents an open toolkit built around 'bridge forms' - identical written words recurring across two or more subject domains with a different sense in each - to test whether transformer language models individuate word occurrences by context beyond the embedding layer. It covers declarative specification of bridge forms, Wikipedia corpus acquisition, occurrence localization, layer-wise representation extraction, domain-pairwise silhouette measurement, and visualization, justifying each choice against failure modes such as sense contamination and subword-tokenization misalignment. It is a methodological and implementation reference and reports no empirical results.

arXiv cs.AI / cs.LG / cs.CL·Sep 4, 16:38 UTC · 11d agoAI research

14

Building and Evaluating Fixed-Voice Thai TTS from Synthetic Speech

Researchers distill a compact 82M-parameter Thai TTS from synthetic OmniVoice data, enabling on-device fixed-voice synthesis without reference audio.

The paper uses a large voice-cloning model (OmniVoice) as a synthetic data source to train Wayu-Paxa-TTS-Edge, an 82M-parameter fixed-voice Thai TTS student. The model achieves 68.2% Challenge-Set Keyword Accuracy (85.5% of Gemini 3.1), 91.4% pause precision, and CERs of 3.7% on Thai and 1.1% on English. It outperforms its teacher on pause placement and is open-sourced with its evaluation framework.

Hugging Face daily papers·Sep 2, 20:00 UTC · 13d agoAI research

25

Continual Learning Mechanisms Compose for Long-Horizon Memorization

Composing data, function, and weight anchors with merged LoRA raises 100-task long-horizon retention from 1.2% to 34.9% in continual fine-tuning.

The paper introduces long-horizon memorization: a model learns 100 query-answer tasks through continual supervised fine-tuning without retaining earlier examples or receiving task identifiers at inference. No single continual learning mechanism maintains strong retention at this horizon, so the authors compose complementary mechanisms along data/function/weight anchors and low-rank allocation rules. The best method combining all three anchors with merged LoRA ranks among the top 3 methods on all three datasets and raises average final retention from 1.2% to 34.9%, a 28-fold improvement.

Hugging Face daily papers·Sep 6, 20:00 UTC · 9d agoAI research

42

Domain-Incremental Learning for Multi-Channel Replay Speech Detection

First continual learning benchmark for multi-channel replay speech detection shows task-specific beamforming cuts catastrophic forgetting across 24 acoustic environments.

Researchers frame replay-attack detection for voice-controlled systems as domain-incremental learning over acoustic environments, evaluating a beamformer-based detector across all 24 environment orderings of the ReMASC corpus with five seeds. Naive sequential fine-tuning raises error rates on previously learned environments by 18.8 points, while elastic weight consolidation halves forgetting but loses plasticity and gradient projection memory is statistically indistinguishable from naive fine-tuning. A task-specific beamformer keeping one spatial front-end per environment significantly improves final and incremental accuracy, and the last environment in a sequence dominates final performance.

arXiv cs.CR·Sep 10, 08:03 UTC · 6d agoResearch1

22

Right Tool, Right Job: Native-Language Evaluation, Tokenizer Sensitivity, and Methodological Findings from a French-Only BabyLM

French BabyLM entry METRON-FR (125M GPT-2, 92.47M words) shows tokenizer artifacts dominate child-scale zero-shot evaluation; proposes standard diagnostics.

METRON-FR is a 125M-parameter GPT-2 pretrained on 92.47M French words, submitted to the BabyLM 2026 Strict track, scoring 85.97% on the native Quebec-French QFrBLiMP benchmark and 62.80% on the BabyLM-weighted leaderboard. A cross-lingual GLUE protocol combining French task-data translation with rank-16 LoRA shows relational tasks gain while world-knowledge tasks regress. Bilingual Lexicon Induction reaches p@1 of 68.84%, 18x above chance, and ablations show single-token zero-shot scoring is dominated by tokenizer and template artifacts at child scale.

arXiv cs.AI / cs.LG / cs.CL·Sep 15, 16:43 UTC · 19h agoAI research

12

Continue, Adapt, or Yield: In-Turn Adaptation to Overlapping Speech in Full-Duplex Agents

Duplex Cue evaluation shows PersonaPlex full-duplex agents adapt in-turn to listener contributions in only 34.8% of collaborative cases versus 68.2% for humans.

The paper introduces Duplex Cue, an evaluation of in-turn adaptation in full-duplex voice agents that separates listener intent (backchannel, collaboration, interruption) from speaker behavior (continue, adapt, yield). Using 208 scorable pairs from 300 human-confirmed cues in unscripted English conversations, it compares recorded human responses with PersonaPlex continuations generated while replaying listener audio. Humans adapt within the turn in 68.2% of collaborative pairs versus 34.8% for PersonaPlex, which otherwise continues unchanged (42.4%) or yields (22.7%).

arXiv cs.AI / cs.LG / cs.CL·Sep 11, 17:46 UTC · 4d agoAI research

15

Reference-Based Bias Detection in LLMs via Relative Representations of Hidden States

Researchers propose auditing LLM bias via relative hidden-state representations, detecting bias increases with 3-50x less compute than output-level benchmarks.

The paper introduces a reference-based bias auditing method that compares hidden-state representations across model variants, such as before and after fine-tuning, by encoding sentences relative to a fixed anchor set. The resulting Representational Bias Shift (Delta-B) correlates with output-level bias change in 15 of 18 tested settings, reaching |r| = 0.84 under full fine-tuning across WildGuardMix, DecodingTrust, and ToxiGen benchmarks. Thresholding Delta-B detects checkpoints whose bias increased with ROC AUC between 0.65 and 0.99 and beats a SEAT-based baseline, while auditing a model in about three minutes with 3-50x less compute.

Hugging Face daily papers·Sep 8, 20:00 UTC · 7d agoAI research

42

MeClear: Cooperative Game-Theoretic Attribution and Risk-Aware Memory Clearance for Long-Horizon LLM Agents

MeClear uses cooperative Shapley attribution to clear harmful memories from long-horizon LLM agents, boosting task recovery by 25.5 points over baselines.

The paper introduces MeClear, a task-conditioned memory clearance framework for long-horizon LLM agents that identifies and selectively suppresses memories with negative downstream utility without permanently altering the persistent memory bank. It combines Leave-One-Out screening with sampled cooperative Shapley attribution to distribute utility across interacting evidence, resolving redundant conflict masking that single-removal evaluations miss. Across ten long dialogue memory pools it achieves 85.9% target recall and 82.3% overall task recovery, a 25.5 percentage-point improvement over LOO baselines.

arXiv cs.AI / cs.LG / cs.CL·Sep 8, 17:46 UTC · 7d agoAI research

32

Training and Finetuning Multi-Vector Embedding Models with Sentence Transformers

Hugging Face published a tutorial on training and finetuning multi-vector embedding models using the Sentence Transformers library.

Hugging Face's blog walks through training and finetuning multi-vector embedding models with Sentence Transformers. Multi-vector approaches store multiple vectors per document to support late-interaction retrieval. The post is a practical guide for developers building retrieval pipelines with the library.

Hugging Face Blog·Aug 26, 00:00 UTC · 21d agoAI tools & infra1

20

ToolLoop: Closed-Loop Tool-Use Data Synthesis via Decomposed Generation and Dynamic Self-Feedback

ToolLoop introduces a closed-loop synthetic data framework whose 11K examples lift a 4B model to 86.40% on BFCL tool-use evaluation.

ToolLoop decomposes tool-use data synthesis into function-name sampling, backward derivation of user queries, and forward derivation of tool calls, with dynamic self-feedback at each stage. This shifts the paradigm from generate-then-filter to generate-verify-refine, reducing inefficient and imbalanced synthetic data. A 4B model trained on 11K synthetic examples reaches 86.40% accuracy on BFCL non-reasoning mode (86.07% in an Isolate variant excluding BFCL-overlapping functions) and 72.1% on ACEBench using only 18.3% of baseline training data.

arXiv cs.AI / cs.LG / cs.CL·Sep 8, 17:26 UTC · 7d agoAI research1

30

Mind2Dialogue: Training Human-Aware Language Models by Simulating User Mental States

Mind2Dialogue simulates users' mental states to generate privileged supervision, boosting personalization and preference-following in Qwen, Llama, and OLMo assistants.

The Mind2Dialogue framework uses a psychology-guided simulator that preserves personal characteristics while updating user mental states through interaction, driving coherent conversations and an Oracle assistant's responses. Privileged distillation trains models on the Oracle's well-informed responses so they can assist users without direct access to mental states at deployment. Training on the full corpus improves every reported personalization metric over Qwen, Llama, and OLMo instruction-tuned baselines, including 26.6 to 40.9 percentage point gains in preference-following generation.

Hugging Face daily papers·updated Sep 14, 17:55 UTC · 1d ago·first Sep 13, 20:00 UTC · 2d agoAI research 2 sources

45

Ask Before You Optimize: Dynamic Pre-Formulation Clarification for Interactive Optimization

Researchers release OR-Clarify, a benchmark testing whether LLM agents ask clarifying questions before formulating optimization models from incomplete requests.

OR-Clarify evaluates pre-formulation clarification in operations research: each task gives a partial problem description, withholds structured hidden slots, and scores agents via bounded interaction with a simulated user, measuring slot recovery, stopping behavior, silent assumptions, and interaction cost. The authors also propose InterOPT, a two-stage framework that identifies formulation-critical gaps to decide when to ask or stop. In choice-based experiments InterOPT substantially outperforms all baselines in exact slot recovery and remains competitive in the open-ended setting.

Hugging Face daily papers·Sep 3, 20:00 UTC · 12d agoAI research1

35

Understanding the Impact of Model Pruning on Long-Tail Forgetting and Explanation Reliability in Medical Imaging

Systematic study finds model pruning causes frequency-dependent long-tail forgetting in medical imaging and that gradient-informed methods best preserve explanations.

Across two long-tailed medical imaging datasets, two CNN architectures, four pruning methods, and sparsity up to 95%, the study measures predictive performance, explanation stability, and faithfulness. Rare classes degrade earlier and more severely than frequent ones, while explanation reliability depends mainly on the pruning strategy, with gradient-informed methods degrading least. Mechanistic analysis ties explanation collapse to loss of class-discriminative gradients rather than vanishing feature activations, recommending class- and explanation-aware evaluation of compression.

arXiv cs.AI / cs.LG / cs.CL·Sep 7, 17:44 UTC · 8d agoAI research

27

The Price of Sparsity: Sufficient Conditions for Sparse Recovery using Sparse and Sparsified Measurements

Researchers derive sufficient sample-size conditions for recovering sparse binary signals from sparse Gaussian measurements, quantifying an information-theoretic threshold of order slog(p/s)/log(ds/p).

The paper studies support recovery of sparse binary signals from noisy linear measurements. For sparse Gaussian designs, the authors identify sufficient minimal sample sizes for maximum-likelihood recovery in the high-SNR regime d*s/p -> infinity, yielding an information-theoretic threshold of order slog(p/s)/log(ds/p) that makes the price of measurement sparsity explicit. They also show a regime where the sample-complexity loss from sparsity is only logarithmic while computational gains are nearly linear, and prove that for independently sparsified dense Gaussian designs a sample size of order p/ψ² suffices for support recovery at any fixed error level.

Hugging Face daily papers·Sep 7, 20:00 UTC · 8d agoAI research

12

Competence-Gated Pooling of Language Models and Priors for Event Forecasting

Paper proposes a competence gate pooling language model forecasts with external priors, improving Brier score from 0.0771 to 0.0732 across 2,357 binary questions.

The paper defines a language model's relative competence as its marginal value beyond an available external forecast, and derives conditions under Brier loss where model disagreement improves that forecast. A competence gate estimates domain-level source weights from resolved outcomes, shrinks uncertain estimates toward a global weight, and recalibrates the pooled forecast. Across 2,357 resolved binary questions and five language models, it improves the external baseline from 0.0771 to 0.0732 Brier and beats global forecast combinations, though it defers to the market on ForecastBench. Across four Qwen models, verbal confidence failed to identify when the model outperformed the external forecast, while outcome-estimated competence supported better abstention.

Hugging Face daily papers·Sep 9, 20:00 UTC · 6d agoAI research

38

SAS: Simple Attention Sparsification via End-to-End Optimization of Context Ranking

SAS trains attention sparsification end-to-end with the language modeling loss, beating sparse attention baselines especially under tight context budgets.

Simple Attention Sparsification (SAS) injects the selector's continuous scores into attention logits in log form inside the softmax, letting gradients from the language modeling loss directly update the ranking of context units. The method uses normalized softmax gates calibrated against the current block and a memory-efficient Triton kernel integrated into FlashAttention-style computation. Across reasoning, long-context, and agentic tasks, SAS consistently outperforms trainable sparse attention baselines across budgets, with the largest gains under tight attention budgets.

Hugging Face daily papers·updated Sep 11, 17:58 UTC · 4d ago·first Sep 10, 20:00 UTC · 5d agoAI research 2 sources1

30