overfeed.news
Anuncio

NexTalk

Automatiza la primera respuesta. Mantén el toque humano.

Chat en vivo con agentes de IA en tu modelo y tu clave de API, con traspaso a tu equipo.

Empezar gratis

Tema

LLMs

3009documentos

250últimos 7 días

arXiv AI Papers

Where Do LLM Ensemble Gains Come From? Decomposing Candidate Value, Diversity, and Aggregation

LLM ensemble gains are often measured against a single baseline, but the value attributed to additional candidates can change with the comparison used. We study an LLM aggregator under four matched conditions: Direct (no candidates), Single-slot (one candidate in a multi-slot interface), Duplicate (one candidate repeated), and Multi-candidate (separately generated candidates). Under either non-diverse control, overall candidate gain (CIV) splits algebraically into an additional-candidate contrast (DSV) and a candidate-context contrast (CCV); the split, but not CIV, depends on the chosen control. On MultiHopRAG model diversity (k=2), the repetition gap Rep=-8.33 pp [-13.25, -3.75] quantifies the difference between gains measured against Duplicate and Single-slot. On a matched FRAMES stronger-panel extension (n=200, k=3, aggregator output limit 4096), paired Delta Rep=+9.67 pp [+6.00, +13.50]; within panels, Rep changes sign (original -6.33; stronger +3.33). In FRAMES evidence aggregation, candidate-context contrasts remain strongly negative under both non-diverse controls. Our results show why ensemble evaluations should state the reference control before attributing gains to additional candidates.

en
arXiv AI Papers

BLADE: Auditing Information Use for Failure Attribution in LLM Multi-Agent Systems

Large language model multi-agent systems extend individual agents' capabilities through collaboration across specialized roles, tools, and workflows. Yet collaboration makes failures harder to diagnose: one agent may misinterpret a correct result while others carry the mistake forward. Failure attribution therefore seeks to identify the responsible agent and decisive error step. Recent methods trace dependencies, but knowing where information came from is not the same as knowing where it went wrong. We introduce BLADE (Blame Localization via Audited Dependency Evidence), a training-free framework that moves beyond tracing information to auditing its use. Using Jev for structured probabilistic judgments, BLADE uses dependency probabilities to select relevant source-decision handoffs for auditing. At these handoffs, it examines whether receiving decisions use their sources faithfully and within task requirements, distinguishing error introduction, propagation, and repair. These findings inform a global diagnosis that considers all execution steps. Without reference answers, BLADE achieves the best or tied-best step accuracy among eight strategies under a shared Jev decision interface on 1,228 failed runs across five Who&When and Who&When Pro subsets, outperforming All-at-once by 2.2-11.1 percentage points. Making responsibility evidence explicit offers a path toward more interpretable and reliable agentic systems.

en
arXiv AI Papers

Looking Complete Is Not Being Grounded: Evaluating LLM Sitreps Against Human-Authored Situation Reports

Evaluation for automatic crisis reporting should measure what practitioners need, not what is convenient to compare. Through LLM-assisted, author-approved content analysis, we derive a 15-slot information schema from professional humanitarian situation reports using qualitative content analysis, then evaluate 121 human and 261 LLM-generated reports across five disasters and three hazard types. Under google/gemini-3.7-flash on Cyclone Idai, the best machine arm is nominally above the empirical human ceiling (0.827 vs. 0.821), but the difference is statistically indistinguishable from zero (95\% bootstrap CI [-.049,.059]; permutation p=.852), and the ordering does not replicate under openai/gpt-5.6-luna. On a blind 100-item Idai sample, the author agrees with the Gemini judge at 0.910 accuracy and unweighted Cohen's κ=.799. All four judges clear the plan's raw-agreement criterion of .7; only Gemini reaches .7 on chance-corrected κ. Across the three primary events, machine reports reproduce 5.4-11.0% of same-day official figures. Across those same events, raw exact-number matching links 14.3-19.6% of official figures and 85.4-90.1% of machine figures to the social stream; manual audit confirms 45/50 declared absences, 4/25 official matches, and 24/25 machine matches semantically. The issuing-provenance gap is positive in all 12 observed event-judge cells: report date, organisation, series, and authority were not supplied at professional-report coverage in any cell. On the three primary events, ROUGE-L is uninformative, BERTScore is negatively associated with completeness after multiplicity correction, and a generic LLM judge changes sign.

en
arXiv AI Papers

Beyond the Verdict: Agentic Code Judging with Verifiable Feedback

Test-time scaling for large language models (LLMs) can involve generating multiple candidate solutions for selection or revising an earlier attempt. To guide selection and revision, we need a verifier that can estimate candidate correctness and provide evidence of failure. In competitive programming, however, official judging systems provide authoritative verdicts under finite submission budgets, without revealing failure explanations or failing inputs. To provide informative feedback without consuming official submissions, we formulate a task in which a model judges and diagnoses a candidate program using only the problem statement and source code, with generated counterexamples serving as checkable evidence. To evaluate this task, we build USACO-JUDGE from the USACO 2026 contests, using accepted and unaccepted programs written by human experts and multiple LLMs. We then train ARBITER, a 35B mixture-of-experts (MoE) agentic verifier, with reinforcement learning in a generic execution sandbox. Training raises counterexample success on incorrect programs from 40.61% to 62.67% and mean binary verdict accuracy on USACO-JUDGE from 71.41% to 87.41%, approaching frontier verification performance. Beyond verification, ARBITER outperforms baselines in both code selection and refinement. In an IOI 2026 case study, ARBITER-guided refinement raises GPT-OSS-120B's expected score above the gold-medal threshold under the official submission limit.

en
arXiv AI Papers

SolBehave: Evaluating the Behavioural Correctness of LLM-Generated Smart Contracts

Large language models increasingly generate Solidity smart contracts that, once deployed, can never be patched, making a single latent fault permanent and exploitable. Yet current evaluations mostly ask only whether generated code compiles. We introduce SolBehave, a benchmark for evaluating the correctness of model-generated contracts at scale, moving beyond ad hoc unit tests. Across 250 real, reproducible Ethereum contracts, we apply three complementary tests of behavioural fidelity: differential fuzzing, historical-transaction replay, and, as our core contribution, synthetic-transaction replay guided by a Dynamic Condition Response (DCR) graph that enumerates the space of legitimate function-call sequences. Differential fuzzing compares each generated function with the original on automatically generated inputs; historical replay compares the whole contract on the transactions it actually received on-chain; DCR-guided replay compares it on sequences of calls that follow its intended usage, including orderings that never occurred on-chain. We evaluate seven state-of-the-art open-weight models after one round of automated repair. Compilation success ranges from 26.4% to 92.7%, but under differential fuzzing only 31.4% of all generated function bodies that compile, pooled over models and samples, are equivalent to the original. Divergence at the contract level is the norm rather than the exception: for six of the seven models, between 13% and 31% of the generated contracts that compile diverge from the original under historical replay, and between 13% and 44% under DCR-guided replay; only Gemma-4-31B stays below 3% on both. Compilation is therefore a necessary condition for correctness and far from a sufficient one.

en
arXiv AI Papers

SP^2ec: Adaptive Self-Speculative Decoding for Vision-Language Models

Speculative Decoding (SD) has demonstrated its efficacy in accelerating large language model inference, but for Vision-Language Models (VLMs), the diverse input modalities make it difficult for a small trained drafter to achieve consistently high throughput. Self-speculative decoding reuses parts of the target model as a drafter which retains the multi-modal capability of the VLMs, yet its efficiency depends on choosing a suitable draft configuration for each prompt. We propose Single-Peak SPeculative decoding (SP^2ec), a training-free bandit algorithm that adapts the draft configuration during generation. SP^2ec exploits the single-peak throughput pattern observed in existing self-speculative methods, including SWIFT, KnapSpec, SpecVLM, and our simple cosSim. Our algorithmic design explicitly minimizes a novel notion of wall-time regret for each given query, corresponding to the empirical wall-time latency. Theoretical guarantees on the wall-time regret show that SP^2ec can exploit the optimal draft configuration, with negligible exploration cost. Extensive experiments on Qwen2.5-VL-7B-Instruct and LLaVA-NeXT-Mistral-7B across visual and textual tasks show consistent improvements of up to 11\% over the original self-speculative methods, demonstrating the wide applicability of the proposed SP^2ec algorithm.

en
arXiv AI Papers

On the Monitorability of Natural Language Autoencoders: Identification vs Discrimination

Natural Language Autoencoders (NLAs) are proposed as oversight tools for language models: an Activation Verbaliser (AV) writes an English explanation of an activation vector, and an Activation Reconstructor (AR) maps the explanation back to a vector. Their value for oversight depends on a human monitor being able to tell a true account of an activation from a false one, given the Av's explanation. We formalise NLA training and NLA oversight deployment as two different referential games from Emergent Communication: training is a generative referential game, while oversight is a discriminative, multi-modal, partially-observable referential game played over a many-to-many relation between the monitored language model's activations and the accounts that are true of them. The gap between the two games predicts that AV's explanations will identify which activation they describe without discriminating its true account from closely matched false ones, since training never asks explanations to rule anything out. Using the NLA's own AR as a maximally-sympathetic surrogate monitor model, we experimentally confirm this theoretical insight on one released checkpoint: true accounts beat accounts of other activations in all 320 comparisons, but beat fluent rivals differing in a single claim only 60.2% of the time (std.err. 6.7%, chance at 50%, n = 35). A meaning-preserving rewrite by an uninvolved model costs only 2.1% of the distance to another activation's account, so the failure is not due to the NLA pair having developed private conventions in the AV's explanations. Our rival accounts test only transcript-level truths, which the monitored language model's conversation settles, so these results bound the easiest case and leave claims about the monitored model's own processing untested. We state conjectures that would make both results consequences of the framework.

en
arXiv AI Papers

AIRRI: Towards a Human-Readable, Machine-Resistant Reading Interface

The rapid advancement of image-to-text extraction systems and large language models (LLMs) poses new challenges for educational assessment by making assignment materials easy to transcribe and solve with minimal human effort. To address this problem, we present AIRRI (Artificial Intelligence Resilient Reading Interface), a framework for generating machine-resistant educational content, a dataset of higher-education assignments for evaluation, and an initial evaluation of the human readability of AIRRI machine-resistant text designs. AIRRI employs two categories of machine-resistance techniques. First, AIRRI applies image transformations including character spacing, word spacing, line spacing, character opacity variation, stripe overlays, and image noise. Second, AIRRI applies decision-based adversarial attacks: RayS, SurFree, and ADBA. We evaluate AIRRI against three image-to-text systems: Tesseract, EasyOCR, and the transformer-based GOT-OCR2 model. Our results show that AIRRI substantially degrades automated text extraction performance. AIRRI increased character error rates for EasyOCR (6.9% to 55.9%), Tesseract (2.3% to 100.0%), and GOT-OCR2 (30.7% to 48.1%). Word error rates increased for EasyOCR (24.9% to 96.7%) and Tesseract (10.7% to 100.0%). Among the evaluated decision-based attacks, ADBA consistently produced the largest degradation in image-to-text performance, while RayS and SurFree produced more moderate reductions. EasyOCR and Tesseract were generally more susceptible to both image-based and adversarial perturbations than the transformer-based GOT-OCR2 model. Our findings suggest that AIRRI can significantly reduce the reliability of image-to-text-to-LLM workflows, increasing the effort required for AI-assisted assignment completion.

en
arXiv AI Papers

Unlocking Token Rewards via Training-Free Reward Attribution

In this paper, we propose an extremely efficient, training-free method to extract token-level reward signals directly from an existing deep reward model. Our core idea is to attribute the overall process reward to individual tokens by estimating each token's influence. This influence is defined as the change in the final macroscopic reward (e.g., the process reward) when a token is replaced with a semantically null token. Naively calculating this influence is computationally infeasible, requiring N forward passes through the PRM for an N-token sequence. We overcome this bottleneck by proposing a highly efficient gradient-based estimator. Specifically, we use a first-order Taylor approximation, which simplifies the influence calculation to the inner product of the difference between the token embedding and the null token embedding, and the gradient of the reward with respect to the token embedding. This requires only a single forward and backward pass. The resulting token-level rewards enable standard RL algorithms to perform precise credit assignment without requiring additional reward model training. Experiments on challenging reasoning benchmarks demonstrate that our method substantially improves policy optimization efficiency and enhances the generalization of LLM reasoning capabilities. Our P2T outperforms the outcome reward by +4.9% on MathVista for Qwen2.5-VL-7B-Instruct, and +11.5% on AIME24 for Qwen2.5-Math-7B. The code will be released via https://github.com/stonewst/P2T.

en
arXiv AI Papers

Lost in Implicit Routing: How LLM Agents Fail to Load Relevant Skills at Scale

Reusable skills let LLM agents load task-specific instructions on demand, but they make execution depend on implicit routing over skill metadata. We test whether this design scales by holding SkillsBench tasks and their golden skills fixed while adding irrelevant skills. We find skill routing collapse: agents increasingly load fewer or no skills, even though the relevant skills remain available. With 1,000 irrelevant skills, loaded-skill recall falls from 73.6% to 27.4% and the zero-load rate rises from 9.3% to 51.6%, whereas explicit selection from the same library retains 73.9% recall. Loading other skills, including plausible alternatives to golden skills, is too rare to explain this decline. Masking only the descriptions of irrelevant skills largely restores skill loading, yet with the metadata intact, explicitly asking agents to plan skill use still yields lower recall than preselection. These findings point to semantic interference from descriptions and instruction interference when skill choice competes with task execution. We also evaluate explicit preselection as a practical mitigation, including its coverage and latency trade-offs. Our analysis informs skill protocols and evaluation for agents with growing skill libraries.

en
arXiv AI Papers

The Geometry of Language Model Logits: Generic Structure and Learned Departures

Autoregressive language models produce, at every prediction step, a vocabulary-sized logit vector whose structure determines the conditional next-token distribution. We empirically identify that the ordered vocabulary-head logits follow a remarkably consistent log-linear relation [ z_{(r)} μ- αr. ] across prediction contexts, architectures, and scales, emerging progressively during pretraining. To our knowledge, this is the first systematic analysis of the logit geometry underlying the distributional assumptions used in the prior decoding work. While controlled order-statistic baselines spanning bounded, light-tailed, and heavier-tailed distributions can largely explain the observed log-linear geometry, they do not explain the systematic shift in the slope α: trained language models exhibit α1.4--1.7, sitting well above the initialization baseline (α0.2) . Moreover, α correlates with predictive performance, suggesting that it captures learned structure beyond generic extreme-value effects.

en
arXiv AI Papers

Two For The Price Of One: Curvature-Aware Pruning Preserves Influence Functions Too

Post-training pruning enables efficient deployment, but conventional methods do not explicitly preserve internal-gradient alignment between sparse and dense models. We study novel connections between downstream performance of compressed models and the ability to retain gradient information via influence-function preservation. Central to this connection is a curvature-aware objective that accounts for both layer activations and backpropagated gradients. Despite its promise, this objective remains underexplored in pruning due to a lack of efficient solvers. We introduce CACTUS, a novel ADMM-based optimizer supporting both pruning and sparse-plus-low-rank decompositions of pre-trained weights. Compared to LLM-Surgeon which targets the same objective, CACTUS can achieve over an order-of-magnitude speedup. At 50\% pruning of Qwen3-4B, CACTUS reduces the C4 perplexity gap to the dense model by 45\%, improves compressed-to-dense per-sample gradient cosine similarity on WikiText-2 from 0.670 to 0.755, and runs over 28faster than LLM-Surgeon on substantially weaker hardware. By preserving influence functions, CACTUS supports effective data selection. On Llama-3.2-3B, fine-tuning a fresh dense model on the 5\% of data selected by CACTUS achieves 49.44\% accuracy on BIG-Bench Hard, compared with 47.22\% using ALPS and 49.54\% using dense-model selection. Finally, integrating CACTUS with LoGRA, a recent gradient-projection strategy, on Qwen3-4B with MLP layers pruned to 2:4 sparsity and a rank-64 correction retains a 0.711 cosine alignment with dense LoGRA-projected C4 gradients.

en
arXiv AI Papers

A Global View of How LLMs Value Human Time

Large language models (LLMs) increasingly act as personal assistants; however, little is known about how they capture national-level preferences on a global scale. Given that "time" is a universal and representative concept, we examine LLMs' time preferences by varying only users' country and measuring the compensation LLMs deem appropriate for a 60-minute appointment delay. We then compare LLMs' responses with GDP per capita on Purchasing Power Parity (GDP_PPP), Mean Income, and a literature-based index (Patience Score). Our results across 193 countries and 5 LLMs demonstrate that people's time across the world is evaluated in noticeably different ways among LLMs, as well as varying levels of robustness under prompt variations. LLMs consistently recommend the highest compensation for Europe, but mostly the lowest for Africa. Based on GDP_PPP, LLMs underestimate the value of time for users living in Asia and South America. These findings indicate that geographical patterns are embedded in LLMs and may translate in different treatment in agentic settings.

en
arXiv AI Papers

Towards Safe Autonomy for Computer Use

The safety of LLM-based agents currently rests on alignment training and safety classifiers, neither of which can rule out that a sophisticated indirect prompt injection (IPI) goes through and causes significant harm. Recent work therefore builds agents with information-flow guarantees that rely on structured tool interfaces to interact with the environment. Computer Use Agents (CUAs) act visually on the computer screen, which does not provide a boundary between trusted first-party interfaces and untrusted, potentially attacker-authored content. Actions are equally ambiguous: a click may land on an application toolbar or a button controlled by an attacker. Treating the user as the source of trust and confirming each sensitive decision produces confirmation fatigue, with the risk of missing harmful actions. We introduce SA-CUA, a dual-LLM framework that generates a reliable trust boundary for CUAs by deriving conservative deterministic rules from the accessibility tree for unmasking trusted regions from the privileged LLM's screenshot, keeping trusted and untrusted content separated. Human interaction is only required for proper task specification in the beginning. During the task a robust endorsement channel allows values to pass from the untrusted to the trusted domain by verifying information via trusted channels. This allows the agent to autonomously handle tasks involving untrusted information. SA-CUA maintains much higher utility than CaMeL-CUA while keeping the attack success rate at zero.

en
arXiv AI Papers

How Low Can You Go: Fine-Tuning LLMs with Bayesian Optimization (LlMBO)

Recently, Morris et al. introduced TinyLoRA, a strategy for LLM fine-tuning which introduces arbitrarily few parameters, and demonstrated accuracy improvements using less than 10 parameters rather than the {10,000} typically demanded by LoRA-based methods. However, that study continued to use gradient-based local search methods of the kind deployed for training multi-billion parameter LLMs in the first place. This article is predicated on the observation that, in the low dimensional setting, global optimization methods become tractable, unlocking Large language Model Bayesian Optimization (LlMBO). LlMBO is sample-efficient, gradient-free, memory-light, parallelizes across devices without synchronization, and, unlike policy-gradient methods, can optimize greedy-decoding accuracy directly. We describe numerical experiments comparing LlMBO against GRPO on fine-tuning tasks for a variety of models on a math benchmark suite which show that LlMBO can be a suitable alternative in this regime, matching or exceeding GRPO's fine tuning gains, while using a fraction of the compute and memory.

en
arXiv AI Papers

EgoOmniWorld: Unifying Egocentric Understanding via Ego-Action-World Interactions

Egocentric video understanding spans diverse capabilities, modalities, and evaluation formats, yet these are often studied in isolation despite their tight coupling in real-world interaction. We propose the Ego-Action-World interaction loop as a unifying principle: the wearer perceives the world, forms goals, acts, changes the world, and uses feedback to guide subsequent behavior. Based on this view, we introduce EgoOmniWorld, a benchmark organized around four complementary interaction structures: Node captures ego, action, and world states; Edge captures directed relations among them; Chain captures linked dependencies; and Loop captures feedback-closed interactions. We develop an interaction-first, evidence-grounded data pipeline that extracts structured interactions from egocentric videos, grounds them in visual and auditory evidence, and generates questions according to their interaction structure. Required modalities and single- or multi-turn formats follow naturally from the supporting evidence and dependencies. EgoOmniWorld contains 24,166 QA groups, including 20,284 training and 3,882 test samples. Experiments on representative open- and closed-source multimodal large language models reveal substantial and uneven limitations across interaction structures. Models trained on EgoOmniWorld improve in-domain interaction understanding and generalize to existing egocentric benchmarks. EgoOmniWorld provides a unified foundation for evaluating and advancing egocentric video understanding. Data, code, and models will be released.

en