每天知道
你的领域
出了什么

Scaling image-to-3D generation to ultra-high resolutions requires controlling rapidly growing computational costs without sacrificing fine geometric detail. We present Filigree3D, a sparse latent flow-matching framework that generates 3D geometry from a single image at voxel resolutions up to $2048^3$, with straightforward extensibility to $4096^3$. To make training tractable, we introduce Structure-Aware Sparse Scaling, which combines spatial bounding with alternating local-global attention to constrain token growth while preserving both fine-scale details and long-range structural context. To enhance detail reconstruction, we curate training samples based on their high-resolution geometric gains and inject multi-scale image features into a sparse 3D DiT, effectively coupling structural semantics with fine-grained visual cues. Furthermore, a visibility-aware voxel regularization strategy improves robustness against sparse perturbations and facilitates the completion of unobserved geometry. Under our default configuration, Filigree3D maintains peak GPU memory consumption within practical limits for contemporary hardware, enabling the generation of highly intricate 3D geometry in approximately one minute. Extensive experiments demonstrate that our method yields substantial improvements in overall geometric fidelity and fine-detail preservation compared to existing baselines, validating practical, detail-preserving 3D generation at unprecedented resolutions.

We present e3j, a fast Euclid-equivariance backend for geometric deep learning applications with JAX bindings for GPU and TPU. Leveraging both optimized CUDA and Pallas kernels and algorithmic improvements, the library achieves state-of-the-art throughput and runtime on both forward and backward paths. On a machine learning interatomic potential (MLIP) use case, it outperforms established backends, measuring up to 34% speed-up over cuEquivariance on water box NPT simulation using MACE, while remaining fully open source. E3j achieves over 80% efficiency over the H100 maximum memory bandwidth on tensor product operations, and in many cases more than doubles throughput of message passing convolutions forward compared to previously available backends. In addition, with the release of dedicated Pallas TPU kernel, e3j opens the possibility of large scale equivariant deep learning workloads on TPU architectures, which has so far been difficult to achieve. Our benchmarks show that e3j also achieves over 80% of a TPUv6e memory bandwidth, up to one order of magnitude more than e3nn-jax. The library is available on GitHub, PyPI and is released under an open source Apache 2.0 license.

Fully homomorphic encryption (FHE) enables neural network inference directly on encrypted inputs, but it remains orders of magnitude slower than plaintext in- ference. Applying the server's plaintext weights to encrypted activations involves plaintext-ciphertext multiplications (PMult) and accounts for more than half of inference time in recent systems. Ternary quantization can replace these multipli- cations with additions and subtractions, but the savings rarely materialize under packed execution. A single PMult applies a weight group fixed by the packing layout and can be avoided only when all its weights share the same ternary value. Ternarizing all groups, however, largely degrades accuracy. We present FIONA, an offline optimizer that selectively ternarizes weights within a given packing layout based on the estimated effect of ternary conversion on the model's performance. FIONA encourages a shared ternary value within each weight group and retains full-precision weights for sensitive groups, so ternar- ized and full-precision paths coexist within a layer. It then compiles these hybrid operators exactly, applying common scaling factors once to accumulated inputs and reusing sums across outputs. Weight ternarization can also narrow the input ranges of downstream polynomials. FIONA fits lower-degree replacements under a cumulative accuracy budget, reducing multiplicative depth and bootstrapping. On VGG11, ViT, and BERT, FIONA reduces PMult operations by 53.4-79.5% and accelerates end-to-end encrypted inference by 2.38x, 1.68x, and 1.84x, re- spectively, with less than 1% accuracy loss across all three models.

Residual completion augments query-aware sparse attention by estimating the contribution of tokens omitted from the exact sparse computation. We ask whether improving a layer's attention-output reconstruction on the same incoming Q/K/V and selected support necessarily improves the fidelity of the final model output. We study training-free RESA and learned Top-K+$φ$ with frozen backbone language models. A prespecified single-layer screen yields two Qwen3-0.6B/Multi-LexSum interventions for which direct-runtime measurements show positive prespecified request-aggregate local reconstruction gain but worse final KL fidelity than the corresponding all-abstain Exact Top-K baseline on both discovery and prompt-token-disjoint holdout requests. Exact restoration at the same layer instead improves final fidelity, showing that the reversal is specific to approximate completion in these cases. In complementary multi-layer experiments, a task-independent local diagnostic often repairs the tested completion estimators, although the repaired models do not consistently outperform Exact Top-K. Together, these results show that better local reconstruction need not translate into better final-model fidelity.

Diffusion drafters accelerate speculative decoding by proposing multiple tokens in parallel. Despite recent advances in speculative decoding through sequence-level drafting and verification, existing training objectives remain largely designed around token-level verification. To address this mismatch, we introduce Block Verification-aware loss (BV loss), a training objective designed to maximize the expected acceptance length of a drafted sequence. BV loss is directly derived from the block verification acceptance rule, providing a principled connection between the drafter training objective and the inference-time verification mechanism at the sequence level. Across math, code, and chat benchmarks, BV loss increases the mean number of tokens accepted per verification call under block verification by 13.0--21.0% over cross-entropy loss training for DFlash and DSpark with Qwen3-4B and Qwen3-8B without changing the inference procedure. BV loss also outperforms tokenwise acceptance objectives such as TV loss and LK loss, and its gains extend to token verification and greedy decoding. These results demonstrate the benefit of training block diffusion drafters with an objective aligned with sequence-level verification, rather than optimizing each token independently.

Language models describe some internal states as good and others as bad. But whether models have a stake in them is an open question. Simply asking models is unlikely to be informative. Any answer may be consistent with genuine introspection, superficial pattern-matching, or with fixed scripts learned in character training. We therefore study revealed preference. Rather than asking about a state, we use activation steering to attach a positively or negatively valenced activation pattern to one of two otherwise meaningless 'zones', switch steering off, and then observe which zone the model prefers. A model with a stake in that state should choose accordingly. Across seven open-weight models from five families, this is indeed what we find. First, steering changes the passages models write about each zone, and those words shift later choice. Second, this shift persists when all surface-level tokens are held fixed and only the hidden KV cache differs. Third, the effect also remains when all text is generated without steering and valence is only injected during cache construction. Thus, the hidden state is sufficient to move choice in proportion to the steering dose. Fourth, the dependence of choice on hidden valence is nearly absent in a base model and emerges during DPO, consistent with a link between valence and goal-directed behaviour formed in training. Finally, given tools to self-steer, a model does not tend to induce a positive state, but it regularly removes an imposed negative state. It does so at a dose-dependent rate and significantly more often than it removes random-direction interventions. Overall we demonstrate that valence-related activation patterns leave hidden traces that predictably govern later choices, even when all visible tokens are identical across conditions. Whether these traces are accompanied by any subjective experience relevant to model welfare remains unclear.

Large language model (LLM) agents execute applications through a workflow of inference requests with tool calls and user interactions. Serving these applications at production scale requires understanding how application behavior shapes inference demand and for guiding efficient execution. Recent characterization studies provide request-level workload measurements and agent execution analysis. However, an end-to-end view connecting task initiation, workflow execution, and inference infrastructure remains unexplored. In this paper, we analyze a two-week trace of 11.7 million requests from a large-scale production platform for general-purpose agents, backed by inference infrastructure comprising over 10k GPUs. We characterize the platform at three connected levels: task-level initiation semantics, workflow-level execution patterns, and infrastructure level serving demands. Our measurements reveal workload patterns such as highly skewed request volumes across sessions, rare execution overlap among logical sibling requests, and context reuse across task boundaries. Building on these observations, we analyze deployment implications and identify open problems to guide future research on agent serving systems.

MeMo (Zanzotto et al., 2025) is a recent language-model architecture that stores associations between token contexts and next tokens in a correlation matrix memory. In this work, we study its single-layer form and show that its ideal retrieval rule is a multiclass classifier based on the positional Hamming kernel. The MeMo architecture represents both the sequence features and the output labels with Gaussian random codes. Its score is therefore a doubly randomized sketch of the ideal classifier. Under independent input and output codebooks, we bound the errors introduced by context sketching and output decoding, characterize their dependence on model and data parameters, and give a margin-based guarantee for recovering the ideal prediction. Controlled simulations support the trends predicted by the analysis. On a restricted WikiText-2 next-token task, we compare single-layer MeMo with classical baselines and show that it can offer a useful trade-off among predictive accuracy, memory, and throughput, particularly on a GPU, where its matrix operations can be parallelized.

Reinforcement learning (RL) post-training often uses distinct GPU kernels for rollout and policy update. In synchronous PPO and GRPO, numerical disagreement can perturb ratios between current token probabilities and those assigned during rollout. Recomputing rollout log-probabilities with the policy-update backend avoids this discrepancy but adds a forward pass. Bitwise-consistent unified kernels permit reuse when the policy snapshot and probability processing match the objective. Their optimization must preserve agreement across distinct execution regimes. We present KernelBraid, an agentic framework starting from a hand-tuned, bitwise-consistent implementation. Its optimization intermediate representation (IR) organizes source-code search by linking implementations and modifications to numerical requirements, workload measurements, and derivation history. The agent coordinates changes and retains verified intermediates for further exploration; promotion requires passing correctness checks and improving aggregate latency within per-workload limits. Across 12 end-to-end training configurations on H20, KernelBraid achieves 1.10x average throughput relative to AReaL with log-probability recomputation, and the mean training-reward ratio rounds to 1.00x. Isolated-layer profiling yields 1.40x average speedup in summed phase time across 15 model-GPU pairs. Operator-level evaluation covers correctness and performance for 10 operators on A100, H20, and H200, all passing the prescribed bitwise checks. Unified-attention search achieves 2.52x speedup in summed workload latency over the starting implementation using 7M LLM tokens; ablations assess the contributions of retained evidence and branch exploration to search efficiency and attained performance. Our code is open-sourced at https://github.com/areal-project/AReaL-TIK.

Post-training quantization (PTQ) enables efficient deployment of large vision-language models (LVLMs), but is typically calibrated on a small set while expected to generalize across diverse downstream tasks. Although recent PTQ methods for LVLMs incorporate sensitivity signals, they still minimize reconstruction loss with respect to the full-precision model, potentially over-preserving FP behavior and calibration-specific bias. Rather than treating quantization solely as an error to be minimized, we observe that it can also provide beneficial regularization for certain layers and modalities. Motivated by this observation, we propose Balanced Fitting, a quantization effect-based framework that balances precision and regularization beyond reconstruction-based optimization. By measuring layer- and component-wise quantization effects for weights, vision activations, and text activations, Balanced Fitting combines fine-grained fitting for sensitive components with coarser fitting to exploit potential regularization benefits. Experiments on multiple LVLMs show that our method consistently outperforms prior PTQ approaches under both weight-only and weight-activation quantization, while lower reconstruction loss does not reliably translate into better downstream performance. The source code is publicly available at https://github.com/kmc3661/BFQ

AlphaFold2 is written in JAX, so the same inference code compiles and runs unchanged on CPUs, GPUs and Google Cloud TPUs. That portability makes the accelerator look like the main decision a user has to make. We show that it is not. Running one AlphaFold2 inference workload across a Colab CPU runtime, an NVIDIA T4 GPU and a dedicated eight-chip Cloud TPU v5e slice, we find a large hardware advantage for the TPU, 0.47 s per call in steady state on a single chip against 13.1 s on the T4 in the same measurement campaign, and three ways in which the software layer decides how much of it a user actually gets. The default execution path uses one chip of the eight, and at list prices the idle capacity makes the slice cost about as much per prediction as the GPU. Batching with JAX's vmap never exceeds single-query throughput, while mapping queries across chips with JAX's pmap gives eight chips 6.5-7.9x the throughput of one on a matched grid; automatic sharding leaves the per-chip footprint unchanged, consistent with replication, most plausibly because AlphaFold2 carries no sharding annotations. Our retained trace analysis of a first call at a new input shape reports about three quarters of the traced span in JAX tracing and compilation rather than execution. Reruns five weeks later reproduced neither cloud baseline, the GPU one off by roughly a factor of two, so the hardware ratio above is specific to one campaign.

The inference efficiency of Multimodal Large Language Models (MLLMs) is severely constrained by massive visual token sequences induced by high-resolution inputs, with computational cost scaling quadratically. Existing approaches primarily focus on downstream token compression, while overlooking a fundamental upstream inefficiency: input resolution is treated as a static, task-agnostic hyperparameter. We propose Task-Conditioned Resolution Routing (TCRR), which formulates visual compression as a task-conditioned decision and employs a lightweight cross-modal router that conditions backbone visual representations on textual semantics via feature-wise modulation and cross-attention to predict the minimal sufficient compression level per query. To support this, we curate a dataset of 500k samples across 12 task categories, labeled via a teacher-oracle pipeline to approximate Pareto-optimal compression scales. Extensive experiments across diverse architectures show that TCRR achieves a superior efficiency frontier, specifically reducing visual FLOPs by 40.9% and latency by 53.7% on Qwen3-VL-8B while preserving competitive performance. Further analysis of scaling behavior confirms that dynamically routing visual compression enables optimal resource allocation without modifying the MLLM backbone.

Some attention heads learn similar patterns across inputs. Reusing these patterns could reduce training cost by avoiding repeated query-key score computation and softmax. Through controlled pretraining comparisons, we identify Selective Attention Freezing (SAF), which selects heads with low attention-pattern variance and replaces their attention weights with fitted post-softmax means halfway through training. We represent these fixed patterns with absolute-position and relative-distance preferences, reducing storage from quadratic to linear in sequence length. A fused kernel reconstructs the patterns and executes ordinary-attention and replaced heads together. At matched training-token budgets, replacing 25% of attention heads gives 1.056x faster post-replacement optimiser updates at 124M parameters and 4K context, with a 0.77% perplexity increase. At 1B and 8K context, post-replacement updates are 1.068x faster on four GPUs including communication, with a 0.51% perplexity increase. The resulting models also accelerate long-input finetuning and causal prefill. After associative-recall adaptation, the 124M model with 25% replacement generalises to more key-value pairs at a fixed length better than ordinary attention and two pruning controls.

Multi-turn rollout dominates the cost of agentic reinforcement learning (RL). Asynchronous execution and elastic GPU resources can accelerate this stage, but adding rollout replicas yields diminishing returns while training GPUs remain idle between updates. We observe that effective resource use also depends on the prefill--decode (PD) configuration. Both the choice between colocation and disaggregation and the optimal PD ratio vary with the workload, making resource scaling and PD configuration interdependent. Exploiting this opportunity requires selecting effective configurations and realizing their benefits within transient resource-availability windows despite reconfiguration costs. We present PEARL, an asynchronous agentic RL system that coordinates external resource elasticity, temporary reuse of idle training GPUs, and adaptive PD execution. PEARL maintains a unified GPU--worker--role state and uses runtime profiles to predict rollout batch completion time, accounting for environment-induced reductions in decode concurrency. It selects the PD mode and ratio under the current GPU budget and translates each decision into an incremental transition plan that minimizes worker and role changes. Cost-aware switching and borrowing policies suppress transitions with insufficient expected benefit while ensuring timely return of training GPUs. Our evaluation show that PEARL achieves $2.17$--$2.79\times$ the throughput of fixed-resource ROLL across different LLMs. Compared with RLBoost+, throughput improves by up to approximately 26.9% for Qwen3-8B and 36.3% for Qwen3-30B-A3B.

Video Large Language Models (Video LLMs) have achieved remarkable progress in video understanding, but their inference efficiency is constrained by the large number of visual tokens produced by long videos. Recent video token compression methods increasingly introduce sophisticated strategies for token selection, pruning, and merging. This raises a fundamental question: how much of compression performance can be obtained by simply preserving the structure encoded in the visual representations? We investigate this question with SimpleCluster, a simple and training-free baseline that performs position-aware cross-frame clustering in the visual feature space and represents each cluster using the mean of its original visual features. Extensive experiments across four video understanding benchmarks and three representative Video LLMs show that SimpleCluster achieves competitive or superior performance over recent compression methods across a wide range of token retention ratios, with particularly strong robustness under extremely low retention rates (e.g., 1%). To understand this behavior, we analyze the feature space preserved by different compression methods in terms of local approximation fidelity and global coverage. The results show that stronger downstream performance is consistently associated with better preservation of the original visual feature distribution, especially its global coverage. These findings highlight feature-space preservation as an important consideration for video token compression under highly constrained token budgets. Our code is available at https://github.com/xiaozhang79/SimpleCluster.

Vision Transformers (ViTs) incur quadratic self-attention cost in the number of tokens. Most token-reduction methods adapt token identities within a prescribed layer-wise compression schedule, or search a static mask offline, and thus limit online adaptation of when and how much to prune. We propose DORA (Dynamic Online Reinforcement Agent), which learns an input-adaptive pruning policy itself for frozen ViTs. At each eligible block, a hierarchical actor decides whether to prune, how many tokens to remove, and which tokens to remove from each image's evolving representation. Because early deletions change the states observed by later decisions, DORA formulates pruning as a finite-horizon Markov decision process. Complete-prefix shadow evaluations convert final-prediction fidelity into localized per-step credit, while closed-loop accuracy feedback adjusts the fidelity penalty toward a shared accuracy-drop target. A privileged critic and all shadow computations are training-only. Deployment retains the frozen backbone and a lightweight actor that applies hard deletion and packed variable-length FlashAttention, converting token reduction into measured speedups. On ImageNet-1K with DeiT-Base, DORA reduces FLOPs by 38.4% relative to the uncompressed backbone within one percentage point of accuracy loss. Averaged across four ViT-type backbones at matched accuracy, DORA uses 13.2% fewer FLOPs and achieves 32.4% higher throughput than the corresponding per-backbone baseline means. Under zero-shot transfer to ImageNet-A, these gains widen to 20.3% and 45.6%, respectively.

Spiking neural networks are attractive for low-power speech command recognition, yet their latency has received far less attention than their energy efficiency, and their multi-timestep execution is widely assumed to make them slower than quantized neural networks. This paper challenges the assumption that more local timesteps necessarily imply higher network latency. By overlapping computation across adjacent layers at the timestep level, SNNs may complete execution in less time than comparable bit-serial QNNs. However, this overlap relies on spikes firing on incomplete inputs, and a spike once generated cannot be withdrawn, so its error persists and reduces accuracy. Waiting for more input before firing would seem to improve accuracy at the cost of reduced overlap. Yet we find and prove that this intuition fails at some layers, where even a small increase in waiting can change spike timing and downstream computation, making the network both slower and less accurate. We therefore propose a Pipeline Delay Search method which selects each layer's delay by balancing task-level accuracy gains against added network latency. We then adapt the selected configurations through spike-based quantization-aware training and bounded tuning of firing thresholds and initial membrane potentials. Together, these steps form Falcon, a framework for Fine-grained Analysis of Latency and Controlled firing which systematically analyzes and optimizes SNN latency under a spatial analog compute-in-memory mapping with shared digital engines. We evaluate Falcon on GSCV2 and SSC, achieving competitive accuracies of 96.31 and 83.02 at modeled network-core latencies of 119.64 and 124.00us, respectively. Together, our analysis and results show that SNNs can compute more yet finish faster, and wait longer yet predict worse, highlighting why Falcon matters for both latency and accuracy.

Attention often concentrates on a small subset of tokens in the context, but which subset matters changes from one query to the next. To exploit this changing structure, we introduce SANTA++, a training-free stochastic attention method that uses representative keys for memory-efficient selection without scanning the entire key-value (KV) cache. Cached keys are organized into teams, and the query scores one representative from each team to decide which teams to sample. We compute exact attention scores within the sampled teams and reweight each team's contribution by the inverse of its inclusion probability. This importance sampling correction estimates attention over the full cache, with a sampling budget that lets us trade memory reads for accuracy. Remarkably, with 32 or 64 sampled teams, SANTA++ uses 16% to 22% of dense attention's KV reads and retains 94% to 99% of the dense-attention baseline's scores on LongBench v2 and HELMET's retrieval-augmented generation subset, and 85% to 91% on RULER, with Qwen2.5-7B-Instruct at 32K context. With 31 sampled teams, our GPU implementation delivers a $1.69\times$ attention speedup over the dense FlashAttention baseline at 32K context. By reducing the number of cache entries read, SANTA++ in principle complements architectures with compressed KV representations, such as multi-head latent attention. Our kernels are available at: https://github.com/OPUSLab/santapp-kernel-demo.git.

Diffusion drafting with autoregressive (AR) verification has emerged as a promising paradigm for efficient speculative decoding. Recent self-speculation models, represented by Nemotron-Labs-Diffusion, further simplify the speculative pipeline by unifying drafting and verification within a shared backbone, while enabling longer acceptance lengths. However, the Pareto frontier between aggregate and per-request throughput remains underexplored. At low concurrency, sequential draft-verify execution requires two model forward passes per round, limiting the effective tokens per forward (TPF). By contrast, at high concurrency, longer drafts incur increasingly expensive computation, forcing individual requests to operate under constrained speculation budgets and preventing full exploitation of the full-backbone drafter. Our key observation indicates that drafting and verification exhibit reciprocal predictability. Draft logits can anticipate likely verification mismatches, while recent verification outcomes predict future drafting utility and suitable block sizes. Building on this observation, we introduce Reciprocal Guidance (RecGuide), a runtime draft-verify orchestration framework that adapts speculative decoding to varying serving loads. RecGuide exploits spare compute capacity through verification-overlapped drafting at low concurrency, while dynamically allocating request-specific draft block sizes as the workload becomes increasingly compute-intensive. Experiments across a wide range of concurrency levels demonstrate consistent throughput improvements over vanilla self-speculation, achieving up to $1.8\times$ speedup.

High-performance kernels underpin efficient accelerator execution but require expert tuning and lengthy manual optimization cycles. LLM coding agents promise automation, yet their CUDA knowledge transfers poorly to data-scarce domain-specific architectures (DSAs) such as NPUs, whose execution models and memory hierarchies differ substantially from those of GPUs. To address this transfer gap, post-training methods adapt LLMs to NPU programming but depend on scarce expert data and substantial training compute. Memory-learning agents instead adapt through external memory, but their uniform credit assignment gives adopted and unused experiences the same reward target, potentially biasing subsequent retrieval rankings. Moreover, when learned values guide only retrieval, high-value experiences that generalize across operators must be retrieved repeatedly rather than retained in context, thereby increasing retrieval overhead and weakening cross-task guidance. We therefore present SAGE, a persistent self-improving agent for NPU kernel synthesis. Adoption-Traced Utility estimation (ATU) combines explicit adoption records with kernel evaluation outcomes for adoption-aware credit assignment. Utility-Gated Consolidation (UGC) uses positive utility and repeated adoption across operators to select and abstract reusable rules into a bounded resident context. On NPUKernelBench, SAGE achieves a 95.5% execution rate versus 84.1% for the strongest controlled baseline, with 86.9% of solved operators outperforming torch_npu. With GLM-5.3, SAGE achieves a 43.99x speedup over the torch_npu reference on sparse flash attention. These results show that adoption-aware credit assignment and selective consolidation enable agents to accumulate and reuse hardware-specific knowledge across tasks.

Visual token pruning has been widely studied as a practical approach to reducing the computational cost of large vision-language models. However, it struggles to preserve essential visual information, which can lead to substantial performance degradation. In particular, image-based token selection can overlook task-relevant details, while text-guided token selection may fail to capture the text--visual relationships needed for complex reasoning. We find that applying textual guidance too early can limit its ability to identify answer-relevant visual regions, whereas text-to-visual attention becomes more informative at intermediate decoder depths. This finding motivates our training-free method, which separates early vision-guided pruning from deferred text-guided reselection. We first prune visual tokens using vision-encoder attention, retain additional candidates until the decoder midpoint, and then use text-to-visual attention to determine the final visual-token set. Across eight benchmarks and three models, our method outperforms the best-performing baselines by an average of 11.10 and 16.84 percentage points in performance recovery at 80% and 90% pruning, respectively, with comparable or lower LLM-prefill latency than most baselines. The source code is publicly available at https://github.com/kmc3661/DeFT

3D vision transformers such as VGGT predict camera poses and scene geometry from multi-view images in a single forward pass, but their global attention over all concatenated view tokens dominates computation as the number of views grows. To reduce this cost, SparseVGGT and HeSS sparsify attention at the block level, and both retain blocks with high attention probability. However, we observe that attention probability poorly predicts how much the model's behavior actually changes when a block is removed, and we show that this mismatch is why performance collapses as sparsity increases. In this paper, we propose ReSS (ReSidual-ReStoring Sparse Attention), which recasts block selection from a problem of maximizing the retained attention mass to one of minimizing the drift that sparsification leaves in the residual stream. We introduce a drift score that quantifies how much each block shifts the residual, and, since the drift of a drop set depends on the directions of the contribution vectors rather than on their magnitudes alone, an iterative residual restoration procedure that refines the drop set as a whole. Across three backbones and five datasets, ReSS preserves dense performance better than prior methods at matched sparsity. Two further results support drift as the quantity that governs the cost of sparsification: maximizing drift degrades performance faster than random selection, and plotted against realized drift instead of sparsity, all methods fall approximately onto a single curve. Code is available at https://github.com/libary753/ReSS.

Leech-lattice quantization gives good quality at two bits per weight, but its codebooks hold more than 10^14 points, too many for a lookup table. Our earlier kernel expanded the codes at load time and read 4.804 bits per weight from GPU memory for 2 bits of code. We present Tetra, a new codebook on the same lattice. A 24-weight block still takes 48 bits, most of which index a 64-state trellis of the Golay code and one shared 16 KiB table. The kernel decodes a block with six table loads and two small lookups inside the matrix-vector product, and reads 2.148 bits per weight. For full models, we retrain one scale per matrix row, store the matrices that lose the most as 4-bit integers, and pay for them with 4-bit embedding tables. Our Qwen3-4B, 8B and 14B files hold 2.73, 2.70 and 2.73 bits per parameter over the whole model. They score 63.37, 69.58 and 75.66 on the full MMLU test set, 4.76, 4.21 and 2.46 points below 4-bit AWQ at 5.3 to 6.0 bits per parameter. They generate 113.8, 95.0 and 57.2 tokens per second in our engine. On GSM8K, through the served kernel, they lose 9.63, 4.62 and 3.26 points to FP16. At 4B our file scores 23.6 points above llama.cpp's IQ2_XXS (2.48 bits per parameter). Every number we measured for a table or figure comes from one NVIDIA L40S GPU. We preregistered the main experiments.

Scaling the horizon of agentic LLMs is bottlenecked by the need to fit ever longer context traces in GPU memory. Context compaction has been the most popular mechanism to alleviate this issue, keeping GPU memory constant for a given trace. Unfortunately, most compaction strategies rely on prefilling the LLM context many times over, hindering training throughput. To alleviate this bottleneck and enable efficient trainable compaction, we propose KV-streams, a plug-and-play strategy compatible with any compaction strategy that substantially increases throughput while showing no evidence of hindering performance. KV-streams enable scalable compaction by streaming the KV cache forward rather than flushing it after each compaction. We show that KV-streams enable three different compaction strategies, achieving a 2.6 to 5x wall-clock speedup in training. Beyond efficiency, we find that the streamed KV cache can act as a recurrent state, carrying forward information that has long since disappeared from the context. Specifically, in a controlled setting we show that, contrary to prior work, RL alone is all that is needed for this behavior to emerge. Overall, we show KV-streams to be an efficient and lightweight plug-and-play addition to any post-training pipeline.

Post-training quantization (PTQ) efficiently compresses Vision Transformers (ViTs) without retraining, yet suffers severe accuracy degradation at low bit-widths. Existing optimization-based PTQ methods guide block reconstruction via either soft logits or second-order Hessian proxies. Logit supervision is prone to overfitting on limited calibration data, while Hessian approximations incur structural truncation errors. To address these limitations, we propose GLF-Q, a novel PTQ framework guided by Global-Local Feature alignment. GLF-Q propagates quantized block outputs through downstream full-precision layers to align penultimate-layer representations under local output regularization, providing downstream feature supervision without explicitly approximating the Hessian or using a Taylor expansion. Furthermore, offline Hadamard transformations are introduced with zero runtime overhead to disperse activation outliers across channels, effectively contracting dynamic ranges and reducing quantization errors. Meanwhile, optimizing this loss via a Straight-Through Estimator (STE) achieves rapid convergence, bypassing continuous relaxation rounding formulations such as AdaRound. Extensive experiments across representative ViT architectures demonstrate that GLF-Q with standard uniform quantizers substantially outperforms state-of-the-art methods under 3-bit quantization on image classification. In addition, GLF-Q exhibits strong out-of-domain calibration robustness and achieves speedups under 8-bit GPU deployment.

Post-training quantization reduces the deployment cost of vision-language models (VLMs), but preserving multimodal capabilities at low bit widths remains challenging. Existing methods rely on modality- or token-level gradient statistics, which are susceptible to cross-sample variations in visual-to-textual token ratios and the positions of visual information, limiting statistical stability. Moreover, overly coarse aggregation through absolute values and averaging discards gradient signs and channel-wise differences, limiting the separation of modality-specific sensitivities. In contrast, the channel space provides a shared coordinate system across samples, making it a more natural basis for capturing stable task-sensitive structures. We therefore propose SubRot, a signed gradient subspace calibration method for VLM rotation quantization. Through eigendecomposition of the empirical Fisher matrix of activation gradients, SubRot identifies a sensitive channel subspace with three properties: cross-sample stability, clear sensitivity separation, and consistent signed effects on the autoregressive loss along certain directions. Guided by a local Taylor expansion, SubRot combines signed first-order guidance along sign-stable directions with second-order constraints along the remaining sensitive directions, while retaining MSE for overall reconstruction quality. This objective steers quantization errors toward loss-decreasing directions while controlling their magnitude. Experiments on five VLMs across five benchmarks show consistent average-score improvements over FlatQuant under W4A6 and W4A4, reaching 1.4 percentage points on LLaVA-NeXT-7B. Under W4A4, average accuracy degradation from FP16 remains within 1.4 percentage points across all evaluated models, while LLaVA-v1.5-13B exceeds its FP16 average score by 0.4 percentage points.

Autoregressive large language model inference repeatedly invokes the target model to generate one token at a time, making generation sensitive to GPU memory movement and sequential execution. This study evaluates two-token multi-token prediction (MTP) against autoregressive decoding in a controlled single-request deployment on an NVIDIA A10G GPU. A 360-request benchmark covered plain-text, reasoning-intensive, and tool-calling workloads, while runtime telemetry, Nsight Systems, PyTorch Profiler, and selected Nsight Compute measurements were used to explain the observed performance. MTP increased output throughput by \(1.91\times\) to \(2.19\times\) across all prompts and reduced time to first output by 10.0--14.2%. Median mean acceptance length ranged from 2.370 to 2.595 tokens per verification iteration. Profiling showed that MTP introduced a longer and more complex execution path, including proposal, sampling, attention, gathering, and reduction operations. However, it required 56.4--78.1% fewer executions of the selected repeating CUDA Graph per generated token. The dominant MTP GEMM kernel was not faster than the dominant autoregressive GEMV kernel, and selected instances of both approached the A10G memory-bandwidth limit. These results show that MTP improved inference through amortization: greater token progress reduced repeated GPU execution sufficiently to outweigh the additional speculative-execution cost.

Vision-language navigation (VLN) models perform well but target compute-rich platforms, limiting deployment on memory- and power-constrained robotic edge devices. Compression alone does not establish whether a VLN model fits the memory, latency, and energy budgets of an edge platform while preserving navigation behavior. We introduce EdgeVLN, a runtime-aware, deployment-ready quantized VLN model that closes this gap. EdgeVLN combines a quantized StreamVLN model with Latent Trajectory Termination Extractor (LATTE), a lightweight causal transformer that improves real-time stopping by predicting a Stop Action verifier rank. Both execute through our llama.cpp VLN driver, which reconstructs streaming context and prunes memory tokens on-board. We characterize a pretrained StreamVLN backbone across weight quantization from 8 to 2 bits and multiple inference runtimes to identify a feasible operating point. LATTE reuses backbone hidden states within the budget freed by quantization, requiring neither a second vision encoder nor an additional backbone forward pass. We evaluate six backbone precisions and seven candidate stop heads on BF16 and IQ4 NL across all 1,839 R2R VLN-CE val-unseen episodes. We measure success rate (SR) in simulation and latency, energy, and resident memory on an NVIDIA Jetson Orin NX 16 GB. LATTE achieves our highest SR, 58.02 percent on the deployed 4-bit model, exceeding the BF16 baseline with only 0.013 s additional latency per navigation step. Four-bit formats achieve nearly identical SR, but step energy varies 36.8 times by execution path. Only IQ4 NL under our VLN driver fits the board, using 11.35 GB resident memory while running 20.8 times faster and using 13.3 times less energy than storage-streamed BF16. INT2 collapses. Runtime selection, memory-token pruning, and quantization are essential for efficient edge deployment.

Video diffusion models are rapidly scaling and exhibiting enhanced generation capabilities. Among these recent advancements, MiniMax-H3 stands out as a highly capable, production-level open-source model. However, its 33-billion parameters and multi-step iterative denoising process introduce substantial computational overhead. Consequently, their practical production is hindered by generation latency in the cloud deployment like NVIDIA-GB200, alongside strict memory limits that pose further challenges at the edge device like DGX-Spark. To address these diverse hardware bottlenecks from cloud to edge device, we present a full-stack inference pipeline that integrates efficient algorithmic design with optimized operator implementations. Algorithmically, we introduce a cross-resolution two-stage generation scheduler that exploits the step-wise nature of diffusion: early low-resolution steps rapidly establish the global layout, while later high-resolution steps focus refinements of local and perceptual details. These stages are connected by a learned latent-to-latent mapping module, completely eliminating the computationally expensive VAE decode-reencode cycle for resolution transferring cross different resolutions. For operator implementation, we deploy a Recursive Self-Improvement (RSI) loop that searches kernel fusions and memory layouts, evaluating latency together with numerical agreement. Together, these optimizations deliver up to 30x end-to-end speedup and 20% lower memory: a 5-second 1344x768 video with audio is generated 3.5x faster than real time on an 8xGB200 node, and in under a minute fully memory-resident on a single DGX Spark.

Long-video generation with diffusion transformers (DiTs) produces extremely long token sequences, making attention a dominant inference bottleneck. Dynamic sparse attention reduces computation, but its realized speedup remains limited because irregular query-row execution degrades L2 cache locality and increases HBM traffic. We present WaveAlign, a lightweight, cache-aware query-row reordering framework for dynamic sparse attention. WaveAlign formulates row ordering as an optimization problem and approximates it with two stages. The first stage derives a low-rank SVD representation of sparse-mask rows and groups query rows with similar K/V access patterns, increasing K/V overlap among concurrently scheduled rows. The second stage exploits streaming GPU scheduling by sorting rows within each wave in descending order of their K/V-block counts, so that short rows from the current wave are followed by long rows from the next. This aligns K/V accesses across wave boundaries and enables shared blocks to be reused before eviction. An adaptive skip module avoids unprofitable reordering. By only permuting query and mask rows, WaveAlign preserves sparse-attention semantics and requires no changes to existing methods or backend kernels. Across two GPU architectures, two video DiTs, and four sparse-attention methods, WaveAlign raises the L2 cache hit ratio from 28.48%--36.35% to 79.38%--89.06%, reduces HBM read traffic by up to 92.11%, and achieves up to 1.25x kernel and 1.17x end-to-end generation speedup without quality loss.