EndPrompt: Efficient Long-Context Extension via Terminal Anchoring

Extending the context window of large language models typically requires training on sequences at the target length, incurring quadratic memory and computational costs that make long-context adaptation expensive and difficult to reproduce. We propose EndPrompt, a method that achieves effective context extension using only short training sequences. The core insight is that exposing a model to long-range relative positional distances does not require constructing full-length inputs: we preserve the original short context as an intact first segment and append a brief terminal prompt as a second segment, assigning it positional indices near the target context length. This two-segment construction introduces both local and long-range relative distances within a short physical sequence while maintaining the semantic continuity of the training text--a property absent in chunk-based simulation approaches that split contiguous context. We provide a theoretical analysis grounded in Rotary Position Embedding and the Bernstein inequality, showing that position interpolation induces a rigorous smoothness constraint over the attention function, with shared Transformer parameters further suppressing unstable extrapolation to unobserved intermediate distances. Applied to LLaMA-family models extending the context window from 8K to 64K, EndPrompt achieves an average RULER score of 76.03 and the highest average on LongBench, surpassing LCEG (72.24), LongLoRA (72.95), and full-length fine-tuning (69.23) while requiring substantially less computation. These results demonstrate that long-context generalization can be induced from sparse positional supervision, challenging the prevailing assumption that dense long-sequence training is necessary for reliable context-window extension. The code is available at https://github.com/clx1415926/EndPrompt.

All Circuits Lead to Rome: Rethinking Functional Anisotropy in Circuit and Sheaf Discovery for LLMs

In this paper, we present empirical and theoretical evidence against a central but largely implicit assumption in circuit and sheaf discovery (CSD), which we term the Functional Anisotropy Hypothesis: the idea that functions in large language models (LLMs) are localised to a unique or near-unique internal mechanism. We show that a single LLM task can instead be supported by multiple, structurally distinct circuits or sheaves that are simultaneously faithful, sparse, and complete. To systematically uncover such competing mechanisms, we introduce Overlap-Aware Sheaf Repulsion, a method that augments the CSD objective with an explicit penalty on structural overlap across multiple discovery runs, enabling the discovery of circuits or sheaves with strong task performance but minimal shared structure across a plethora of common CSD benchmarks. We find that this phenomenon becomes increasingly pronounced as the number of discovered sheaves grows and persists robustly across major CSD methods. We further identify an ultra-sparse three-edge sheaf and show that none of its edges is individually indispensable, undermining even weakened notions of canonical or essential components. To explain these findings, we propose a Distributive Dense Circuit Hypothesis and provide a theoretical analysis demonstrating that non-unique, low-overlap circuit explanations arise naturally from high-dimensional superposition under mild assumptions. Together, our results suggest that mechanistic explanations in LLMs are inherently non-canonical and call for a rethinking of how CSD results should be interpreted and evaluated.

What Happens Inside Agent Memory? Circuit Analysis from Emergence to Diagnosis

Agent memory failures are silent: an LLM-based agent can produce a fluent response even when it fails to extract, retain, or retrieve the information needed across sessions. The write-manage-read loop describes the external pipeline of these systems but leaves open which internal computations implement each stage. Tracing internal feature circuits across the Qwen-3 family (0.6B--14B) and two memory frameworks (mem0 and A-MEM), we report three findings. First, control is detectable before content: routing circuitry is causally active at 0.6B, while content circuitry produces no detectable signal until 4B under our tracing setup, creating a deployment regime where small models route with apparent competence but silently fail at extraction and grounding. Second, within the content group, Write and Read share a late-layer hub that operates as a context-grounding substrate already present in the base model; only memory framing recruits a functional grounding direction on this substrate, and the hub transfers across both frameworks. Third, emergence does not imply steerability: although the content circuit becomes detectable at 4B, it becomes reliably steerable only at 8B, indicating that detection and intervention have distinct scale thresholds. As a practical implication, the feature-space separation between the two circuit groups enables per-operation failure localization at 76.2% accuracy without supervision, providing a stage-level diagnostic for otherwise silent agent-memory failures.

Reinforcing Consistency in Video MLLMs with Structured Rewards

Multimodal large language models (MLLMs) have achieved remarkable progress in video understanding. However, seemingly plausible outputs often suffer from poor visual and temporal grounding: a model may fabricate object existence, assign incorrect attributes, or collapse repeated events while still producing a globally reasonable caption or answer. We study this failure mode through a compositional consistency audit that decomposes a caption into supporting factual and temporal claims, investigating whether a correct high-level prediction is actually backed by valid lower-level evidence. Our top-down audit reveals that even correct root relational claims often lack reliable attribute and existence support. This indicates that standard sentence-level supervision is a weak proxy for faithful video understanding. Furthermore, when turning to reinforcement learning (RL) for better alignment, standard sentence-level rewards often prove too coarse to accurately localize specific grounding failures. To address this, we replace generic sentence-level rewards with a structured reward built from factual and temporal units. Our training objective integrates three complementary components: (1) an instance-aware scene-graph reward for factual objects, attributes, and relations; (2) a temporal reward for event ordering and repetition; and (3) a video-grounded VQA reward for hierarchical self-verification. Across temporal, general video understanding, and hallucination-oriented benchmarks, this objective yields consistent gains on open-source backbones. These results suggest that structured reward shaping is a practical route to more faithful video understanding.

CodeHacker: Automated Test Case Generation for Detecting Vulnerabilities in Competitive Programming Solutions

The evaluation of Large Language Models (LLMs) for code generation relies heavily on the quality and robustness of test cases. However, existing benchmarks often lack coverage for subtle corner cases, allowing incorrect solutions to pass. To bridge this gap, we propose CodeHacker, an automated agent framework dedicated to generating targeted adversarial test cases that expose latent vulnerabilities in program submissions. Mimicking the hack mechanism in competitive programming, CodeHacker employs a multi-strategy approach, including stress testing, anti-hash attacks, and logic-specific targeting to break specific code submissions. To ensure the validity and reliability of these attacks, we introduce a Calibration Phase, where the agent iteratively refines its own Validator and Checker via self-generated adversarial probes before evaluating contestant code.Experiments demonstrate that CodeHacker significantly improves the True Negative Rate (TNR) of existing datasets, effectively filtering out incorrect solutions that were previously accepted. Furthermore, generated adversarial cases prove to be superior training data, boosting the performance of RL-trained models on benchmarks like LiveCodeBench.

Not All Preferences Are Created Equal: Stability-Aware and Gradient-Efficient Alignment for Reasoning Models

Preference-based alignment is pivotal for training large reasoning models; however, standard methods like Direct Preference Optimization (DPO) typically treat all preference pairs uniformly, overlooking the evolving utility of training instances. This static approach often leads to inefficient or unstable optimization, as it wastes computation on trivial pairs with negligible gradients and suffers from noise induced by samples near uncertain decision boundaries. Facing these challenges, we propose SAGE (Stability-Aware Gradient Efficiency), a dynamic framework designed to enhance alignment reliability by maximizing the Signal-to-Noise Ratio of policy updates. Concretely, SAGE integrates a coarse-grained curriculum mechanism that refreshes candidate pools based on model competence with a fine-grained, stability-aware scoring function that prioritizes informative, confident errors while filtering out unstable samples. Experiments on multiple mathematical reasoning benchmarks demonstrate that SAGE significantly accelerates convergence and outperforms static baselines, highlighting the critical role of policy-aware, stability-conscious data selection in reasoning alignment.

Mitigating Hallucinations in Video Large Language Models via Spatiotemporal-Semantic Contrastive Decoding

Although Video Large Language Models perform remarkably well across tasks such as video understanding, question answering, and reasoning, they still suffer from the problem of hallucination, which refers to generating outputs that are inconsistent with explicit video content or factual evidence. However, existing decoding methods for mitigating video hallucinations, while considering the spatiotemporal characteristics of videos, mostly rely on heuristic designs. As a result, they fail to precisely capture the root causes of hallucinations and their fine-grained temporal and semantic correlations, leading to limited robustness and generalization in complex scenarios. To more effectively mitigate video hallucinations, we propose a novel decoding strategy termed Spatiotemporal-Semantic Contrastive Decoding. This strategy constructs negative features by deliberately disrupting the spatiotemporal consistency and semantic associations of video features, and suppresses video hallucinations through contrastive decoding against the original video features during inference. Extensive experiments demonstrate that our method not only effectively mitigates the occurrence of hallucinations, but also preserves the general video understanding and reasoning capabilities of the model.

$λ$-GRPO: Unifying the GRPO Frameworks with Learnable Token Preferences

Reinforcement Learning with Human Feedback (RLHF) has been the dominant approach for improving the reasoning capabilities of Large Language Models (LLMs). Recently, Reinforcement Learning with Verifiable Rewards (RLVR) has simplified this paradigm by replacing the reward and value models with rule-based verifiers. A prominent example is Group Relative Policy Optimization (GRPO). However, GRPO inherently suffers from a length bias, since the same advantage is uniformly assigned to all tokens of a response. As a result, longer responses distribute the reward over more tokens and thus contribute disproportionately to gradient updates. Several variants, such as DAPO and Dr. GRPO, modify the token-level aggregation of the loss, yet these methods remain heuristic and offer limited interpretability regarding their implicit token preferences. In this work, we explore the possibility of allowing the model to learn its own token preference during optimization. We unify existing frameworks under a single formulation and introduce a learnable parameter $\lambda$ that adaptively controls token-level weighting. We use $\lambda$-GRPO to denote our method, and we find that $\lambda$-GRPO achieves consistent improvements over vanilla GRPO and DAPO on multiple mathematical reasoning benchmarks. On Qwen2.5 models with 1.5B, 3B, and 7B parameters, $\lambda$-GRPO improves average accuracy by $+1.9\%$, $+1.0\%$, and $+1.7\%$ compared to GRPO, respectively. Importantly, these gains come without any modifications to the training data or additional computational cost, highlighting the effectiveness and practicality of learning token preferences.

Staying in the Sweet Spot: Responsive Reasoning Evolution via Capability-Adaptive Hint Scaffolding

Reinforcement learning with verifiable rewards (RLVR) has achieved remarkable success in enhancing the reasoning capabilities of large language models (LLMs). However, existing RLVR methods often suffer from exploration inefficiency due to mismatches between the training data's difficulty and the model's capability. LLMs fail to discover viable reasoning paths when problems are overly difficult, while learning little new capability when problems are too simple. In this work, we formalize the impact of problem difficulty by quantifying the relationship between loss descent speed and rollout accuracy. Building on this analysis, we propose SEELE, a novel supervision-aided RLVR framework that dynamically adjusts problem difficulty to stay within the high-efficiency region. SEELE augments each training sample by appending a hint (part of a full solution) after the original problem. Unlike previous hint-based approaches, SEELE deliberately and adaptively adjusts the hint length for each problem to achieve an optimal difficulty. To determine the optimal hint length, SEELE employs a multi-round rollout sampling strategy. In each round, it fits an item response theory model to the accuracy-hint pairs collected in preceding rounds to predict the required hint length for the next round. This instance-level, real-time difficulty adjustment aligns problem difficulty with the evolving model capability, thereby improving exploration efficiency. Experimental results show that SEELE outperforms Group Relative Policy Optimization (GRPO) and Supervised Fine-tuning (SFT) by +11.8 and +10.5 points, respectively, and surpasses the best previous supervision-aided approach by +3.6 points on average across six math reasoning benchmarks.

Pretraining on the Test Set Is No Longer All You Need: A Debate-Driven Approach to QA Benchmarks

As frontier language models increasingly saturate standard QA benchmarks, concerns about data contamination, memorization, and escalating dataset creation costs persist. We propose a debate-driven evaluation paradigm that transforms any existing QA dataset into structured adversarial debates--where one model is given the official answer to defend, and another constructs and defends an alternative answer--adjudicated by a judge model blind to the correct solution. By forcing multi-round argumentation, this approach substantially increases difficulty while penalizing shallow memorization, yet reuses QA items to reduce curation overhead. We make two main contributions: (1) an evaluation pipeline to systematically convert QA tasks into debate-based assessments, and (2) a public benchmark that demonstrates our paradigm's effectiveness on a subset of MMLU-Pro questions, complete with standardized protocols and reference models. Empirical results validate the robustness of the method and its effectiveness against data contamination--a Llama 3.1 model fine-tuned on test questions showed dramatic accuracy improvements (50% -> 82%) but performed worse in debates. Results also show that even weaker judges can reliably differentiate stronger debaters, highlighting how debate-based evaluation can scale to future, more capable systems while maintaining a fraction of the cost of creating new benchmarks. Overall, our framework underscores that "pretraining on the test set is no longer all you need," offering a sustainable path for measuring the genuine reasoning ability of advanced language models.