Papers

Field

Top tools here

Coming with more sources: stars history

All tools

NVIDIA Dynamo+60% downloads · 15.8k/wk+60.0%
BentoML+30% downloads · 41k/wk+29.7%
MLX+26% downloads · 342.6k/wk+25.5%
mlx-lm+8% downloads · 140.9k/wk+7.5%
vLLM1 release this week

Top tools here

Open problems in Inference & serving

Efficient long-horizon context management3 papers raise it · 3 address it
Long-horizon agents need to retain and reuse information without exceeding finite context windows or degrading performance.
- REMORY: Learning Residual Memory for Context Compactionraises · addresses
- Pack Light, CarryOn: Building Object-Oriented Context for Long-Horizon Agentsraises · addresses
- MemT-1.0: Towards Memory-Native Evolutionraises · addresses
Token-level LLM routing efficiency1 paper raises it · 1 addresses it
Fine-grained token-level routing improves cost-quality trade-offs but strains existing serving systems with desynchronization and complexity.
- TokenRouter: Efficient Serving System for Token-Level LLM Routingraises · addresses
Memory-induced sycophancy in agents1 paper raises it · 1 addresses it
Persistent memories can cause agents to over-align with user beliefs, even when those beliefs are incorrect or outdated.
- Memadapter: Counterfactual Adaptation Against Memory-induced Sycophancyraises · addresses
Quality gap in parallel token generation2 papers raise it · 2 address it
Diffusion language models generate tokens in parallel but lag behind autoregressive models in quality due to uniform computation allocation.
- ALoDLM: Adaptively Looped Diffusion Language Modelsraises · addresses
- Pivot-SD: Efficient Self-Distillation for Masked Diffusion Language Modelsraises · addresses
Sparse attention computation overhead1 paper raises it · 1 addresses it
Sparse attention reduces computation but still requires scoring all keys, limiting efficiency gains.
- DoSA: Dots Sparse Attentionraises · addresses

Named by an AI from these papers' abstracts.

Open problems

Recently funded

Market

EMode PhotonixPre-seed · Oct 16$1M
MueonSeries A · Oct 15$65M
TrimagnetixSeed · Oct 15$1M
ModalSeries B+ · Oct 14$750M
Salience LabsSeries B+ · Oct 14$125M

Recently funded

Topics in this field

agent memory · memory construction · term memory257 papers52 this week
gpus · gpu · throughput220 papers32 this week
diffusion language models · language models diffusion · models diffusion language150 papers29 this week
continual learning · catastrophic forgetting · forgetting237 papers25 this week
speculative decoding accelerates · self speculative decoding · speculative decoding89 papers24 this week
sparse attention · linear attention · attention layers134 papers23 this week

Topics in this field

Active contributors · 90 days

Active contributors

Activity · 12 weeks

Coming with more sources: repos

Activity

Recent papers

All in Research →

arXiv cs.AIOct 15
Disagreement-Guided Weak-to-Strong Routing
Sima Noorani, Shayan Kiyani, George Pappas, Hamed Hassani
Small language models are increasingly deployed on edge devices and robotic systems to meet cost and latency constraints, yet their limited capabilities can make reliable deployment challenging. Stronger models can provide assistance when needed, but deciding when to route a weak-model response to a stronger model often relies on signals of response correctness. In open-ended and agentic tasks, however, obtaining such signals can require costly ground-truth supervision that is difficult to obtain at scale. We revisit the foundations of weak-to-strong routing and ask whether routing decisions can instead be guided by signals that do not require ground-truth supervision. We first characterize the optimal weak-to-strong routing policy, while allowing for multiple weak-model responses generated through local inference. Building on this characterization, we derive a principled approximation to the optimal routing based on model-to-model comparison. In particular, we introduce a notion of model disagreement that can be measured directly from model-generated responses, without ground-truth supervision. Importantly, the approximation becomes increasingly accurate as the strong model…
PDF arXiv
arXiv cs.LGOct 15
Query-Efficient Rejection Sampling for Parallel Masked Diffusion
Shiyuan Zhang, Xuheng Li, Qiwei Di, Quanquan Gu
Masked diffusion models (MDMs) accelerate generation by unmasking multiple tokens in parallel, but sampling these tokens independently ignores their correlations and biases the output. Anari et al. (2026) corrected this bias with rejection sampling, decomposing the log acceptance probability into a sum of token-level terms, but the evaluation requires one score query per unmasked token. We propose the Single-Increment Rejection Sampler (SIRS), which provides an estimate of the log acceptance probability from a single randomly sampled increment at the cost of one query. SIRS then converts a Poisson number of clipped estimates into accept/reject decisions through a Bernoulli factory adapted from Chen et al. (2026). Since each query returns the conditionals of all remaining tokens in the block, we further Rao-Blackwellize each increment by averaging over these tokens at no extra cost, thereby reducing clipping. We show that the KL-divergence between the data distribution and the SIRS output is bounded by a clipping term plus the model's training error. With exact scores, parallel decoding pays the full within-block dependence, whereas SIRS pays only the part beyond the clipping range.
PDF arXiv
arXiv cs.DCOct 15
Architecting for Hardware Diversity: Meta's End-to-End Co-Design for High-Performance AI Training on AMD GPUs
Haoyu Zhang, Arshad Rahman, Adnan Aziz, Ajith Prabhudev, Amin Farmahini +53
Meta deploys a heterogeneous fleet comprising NVIDIA GPUs, AMD GPUs, and proprietary MTIA accelerators to reduce costs and mitigate vendor lock-in. However, the extensive history of NVIDIA-specific optimizations hinders performance portability to other hardware. This paper presents our methods for optimizing alternative accelerators, using AMD Instinct MI350X versus NVIDIA B200 for recommendation model training as a case study. We propose three reusable principles. First, hardware-model co-design is essential: rather than directly porting models from NVIDIA GPUs, revising model structures and software recipes allows full exploitation of accelerator-specific strengths (e.g., MI350X's 1.6x HBM capacity relative to B200). Second, our analytical performance model enables zero-warmup optimal kernel selection across diverse datatypes and matrix shapes in production models, drastically simplifying tuning. Third, increasingly mature agentic workflows have lowered the barrier to porting from the CUDA ecosystem; we extend these workflows for microarchitecture-aware kernel optimizations and custom shim generation. Compared to the directly ported baseline, these methods deliver up to 1.5x…
PDF arXiv
arXiv cs.CLOct 15
Beyond Downstream QA: Rethinking How We Evaluate Agent Memory
Sebastian-Antonio Toma, Jean de Kernier, Mike Ply, Oliviu Matei
Agent memory benchmarks evaluate whether an agent's responses reflect correct knowledge, measuring downstream question-answering accuracy as a proxy for memory quality. We show this evaluation paradigm is incomplete: it conflates memory correctness with retrieval and generation quality, producing both false positives where correct responses mask broken memory and false negatives where incorrect responses blame correct memory for retrieval failures. We introduce FEB-30, a benchmark of 30 fact-evolution trajectories spanning 6 diagnostic categories and 3 difficulty levels, constructed via paradigm inversion: documents are generated from known fact trajectories, yielding deterministic ground truth without human annotation. Each trajectory defines how a fact should evolve across multi-author artifacts (documents, emails, conversations), enabling direct inspection of the memory store after each ingestion step. Applying FEB-30 to a representative memory system, we demonstrate that intrinsic and extrinsic evaluation diverge by an average of 16 percentage points, with 44% of trajectories exceeding meaningful divergence thresholds. Diagnostic decomposition reveals that the system…
PDF arXiv
arXiv cs.CLOct 15
Hy-Memory-30B-A3B: Attributed Reinforcement Learning for Long-Term Memory Writing
Tianxiang Fei, Jiazheng Kang, Mao Zheng, Xingyan He, Xiang Yu
External-memory systems extend language-model applications beyond a context window, but their quality depends critically on the model executing the write path: deciding what to retain, how precisely to represent it, and how new evidence should update persistent state. Existing frameworks typically place a general-purpose large language model behind framework-specific prompts. This raises serving cost, and a writer specialized for one protocol does not transfer directly to another; a unified model that can operate across established memory frameworks remains largely unexplored. We present Hy-Memory-30B-A3B, a compact long-term-memory writer trained across Mem0, MemOS, and Hy-Memory. Memory On-Policy Distillation teaches framework-specific write protocols on student-generated states; Memory RL then optimizes downstream answerability using dense per-chunk probes, attributed cross-chunk rewards, over-storage controls, and chunk-level GRPO. Carrying policy-written banks across windows keeps training aligned with deployment. On fixed 20% benchmark subsamples, Hy-Memory-30B-A3B ranks first on all four evaluated harnesses, including held-out EverMemOS, with a 55.49% macro average versus…
PDF arXiv
arXiv cs.LGOct 15
LCT: Losslessly Compressed Tensors with GPU friendly execution for Training
Max Zhu, Bohang Sun, Pietro Lio, Jialin Yu
Lossless tensor compression could reduce memory usage during training, but training requires tensors to be repeatedly re-encoded. We introduce Losslessly Compressed Tensors (LCT), a GPU execution framework designed for this repeated encode-decode regime. LCT exploits the natural redundancy in floating point data to reduce storage for weights, activations and optimizer states. Highly optimized GPU kernels implement Huffman coding with precomputed codebooks for low compression and decompression overhead, essential for training. Our analysis characterizes LCT's storage costs and adaptation to magnitude shifts. LCT achieves a compression ratio of up to $1.455\times$, and is demonstrated in training and finetuning settings. On a 162M-parameter NanoGPT, LCT reduces memory usage by $24.6\%$ for a $7.2\%$ step-time increase. In Qwen3-4B fine-tuning, LCT saves up to $23.8\%$ of peak memory, enabling a $3.5\times$ larger context length in training on the same GPU.
PDF arXiv

Papers

Top tools here

Coming with more sources: stars history

All tools

NVIDIA Dynamo+60% downloads · 15.8k/wk+60.0%
BentoML+30% downloads · 41k/wk+29.7%
MLX+26% downloads · 342.6k/wk+25.5%
mlx-lm+8% downloads · 140.9k/wk+7.5%
vLLM1 release this week

Topics in this field

agent memory · memory construction · term memory257 papers52 this week
gpus · gpu · throughput220 papers32 this week
diffusion language models · language models diffusion · models diffusion language150 papers29 this week
continual learning · catastrophic forgetting · forgetting237 papers25 this week
speculative decoding accelerates · self speculative decoding · speculative decoding89 papers24 this week
sparse attention · linear attention · attention layers134 papers23 this week

Recently funded

Market

EMode PhotonixPre-seed · Oct 16$1M
MueonSeries A · Oct 15$65M
TrimagnetixSeed · Oct 15$1M
ModalSeries B+ · Oct 14$750M
Salience LabsSeries B+ · Oct 14$125M

Active contributors · 90 days

Recent papers

All in Research →

arXiv cs.AIOct 15
Disagreement-Guided Weak-to-Strong Routing
Sima Noorani, Shayan Kiyani, George Pappas, Hamed Hassani
Small language models are increasingly deployed on edge devices and robotic systems to meet cost and latency constraints, yet their limited capabilities can make reliable deployment challenging. Stronger models can provide assistance when needed, but deciding when to route a weak-model response to a stronger model often relies on signals of response correctness. In open-ended and agentic tasks, however, obtaining such signals can require costly ground-truth supervision that is difficult to obtain at scale. We revisit the foundations of weak-to-strong routing and ask whether routing decisions can instead be guided by signals that do not require ground-truth supervision. We first characterize the optimal weak-to-strong routing policy, while allowing for multiple weak-model responses generated through local inference. Building on this characterization, we derive a principled approximation to the optimal routing based on model-to-model comparison. In particular, we introduce a notion of model disagreement that can be measured directly from model-generated responses, without ground-truth supervision. Importantly, the approximation becomes increasingly accurate as the strong model…
PDF arXiv
arXiv cs.LGOct 15
Query-Efficient Rejection Sampling for Parallel Masked Diffusion
Shiyuan Zhang, Xuheng Li, Qiwei Di, Quanquan Gu
Masked diffusion models (MDMs) accelerate generation by unmasking multiple tokens in parallel, but sampling these tokens independently ignores their correlations and biases the output. Anari et al. (2026) corrected this bias with rejection sampling, decomposing the log acceptance probability into a sum of token-level terms, but the evaluation requires one score query per unmasked token. We propose the Single-Increment Rejection Sampler (SIRS), which provides an estimate of the log acceptance probability from a single randomly sampled increment at the cost of one query. SIRS then converts a Poisson number of clipped estimates into accept/reject decisions through a Bernoulli factory adapted from Chen et al. (2026). Since each query returns the conditionals of all remaining tokens in the block, we further Rao-Blackwellize each increment by averaging over these tokens at no extra cost, thereby reducing clipping. We show that the KL-divergence between the data distribution and the SIRS output is bounded by a clipping term plus the model's training error. With exact scores, parallel decoding pays the full within-block dependence, whereas SIRS pays only the part beyond the clipping range.
PDF arXiv
arXiv cs.DCOct 15
Architecting for Hardware Diversity: Meta's End-to-End Co-Design for High-Performance AI Training on AMD GPUs
Haoyu Zhang, Arshad Rahman, Adnan Aziz, Ajith Prabhudev, Amin Farmahini +53
Meta deploys a heterogeneous fleet comprising NVIDIA GPUs, AMD GPUs, and proprietary MTIA accelerators to reduce costs and mitigate vendor lock-in. However, the extensive history of NVIDIA-specific optimizations hinders performance portability to other hardware. This paper presents our methods for optimizing alternative accelerators, using AMD Instinct MI350X versus NVIDIA B200 for recommendation model training as a case study. We propose three reusable principles. First, hardware-model co-design is essential: rather than directly porting models from NVIDIA GPUs, revising model structures and software recipes allows full exploitation of accelerator-specific strengths (e.g., MI350X's 1.6x HBM capacity relative to B200). Second, our analytical performance model enables zero-warmup optimal kernel selection across diverse datatypes and matrix shapes in production models, drastically simplifying tuning. Third, increasingly mature agentic workflows have lowered the barrier to porting from the CUDA ecosystem; we extend these workflows for microarchitecture-aware kernel optimizations and custom shim generation. Compared to the directly ported baseline, these methods deliver up to 1.5x…
PDF arXiv
arXiv cs.CLOct 15
Beyond Downstream QA: Rethinking How We Evaluate Agent Memory
Sebastian-Antonio Toma, Jean de Kernier, Mike Ply, Oliviu Matei
Agent memory benchmarks evaluate whether an agent's responses reflect correct knowledge, measuring downstream question-answering accuracy as a proxy for memory quality. We show this evaluation paradigm is incomplete: it conflates memory correctness with retrieval and generation quality, producing both false positives where correct responses mask broken memory and false negatives where incorrect responses blame correct memory for retrieval failures. We introduce FEB-30, a benchmark of 30 fact-evolution trajectories spanning 6 diagnostic categories and 3 difficulty levels, constructed via paradigm inversion: documents are generated from known fact trajectories, yielding deterministic ground truth without human annotation. Each trajectory defines how a fact should evolve across multi-author artifacts (documents, emails, conversations), enabling direct inspection of the memory store after each ingestion step. Applying FEB-30 to a representative memory system, we demonstrate that intrinsic and extrinsic evaluation diverge by an average of 16 percentage points, with 44% of trajectories exceeding meaningful divergence thresholds. Diagnostic decomposition reveals that the system…
PDF arXiv
arXiv cs.CLOct 15
Hy-Memory-30B-A3B: Attributed Reinforcement Learning for Long-Term Memory Writing
Tianxiang Fei, Jiazheng Kang, Mao Zheng, Xingyan He, Xiang Yu
External-memory systems extend language-model applications beyond a context window, but their quality depends critically on the model executing the write path: deciding what to retain, how precisely to represent it, and how new evidence should update persistent state. Existing frameworks typically place a general-purpose large language model behind framework-specific prompts. This raises serving cost, and a writer specialized for one protocol does not transfer directly to another; a unified model that can operate across established memory frameworks remains largely unexplored. We present Hy-Memory-30B-A3B, a compact long-term-memory writer trained across Mem0, MemOS, and Hy-Memory. Memory On-Policy Distillation teaches framework-specific write protocols on student-generated states; Memory RL then optimizes downstream answerability using dense per-chunk probes, attributed cross-chunk rewards, over-storage controls, and chunk-level GRPO. Carrying policy-written banks across windows keeps training aligned with deployment. On fixed 20% benchmark subsamples, Hy-Memory-30B-A3B ranks first on all four evaluated harnesses, including held-out EverMemOS, with a 55.49% macro average versus…
PDF arXiv
arXiv cs.LGOct 15
LCT: Losslessly Compressed Tensors with GPU friendly execution for Training
Max Zhu, Bohang Sun, Pietro Lio, Jialin Yu
Lossless tensor compression could reduce memory usage during training, but training requires tensors to be repeatedly re-encoded. We introduce Losslessly Compressed Tensors (LCT), a GPU execution framework designed for this repeated encode-decode regime. LCT exploits the natural redundancy in floating point data to reduce storage for weights, activations and optimizer states. Highly optimized GPU kernels implement Huffman coding with precomputed codebooks for low compression and decompression overhead, essential for training. Our analysis characterizes LCT's storage costs and adaptation to magnitude shifts. LCT achieves a compression ratio of up to $1.455\times$, and is demonstrated in training and finetuning settings. On a 162M-parameter NanoGPT, LCT reduces memory usage by $24.6\%$ for a $7.2\%$ step-time increase. In Qwen3-4B fine-tuning, LCT saves up to $23.8\%$ of peak memory, enabling a $3.5\times$ larger context length in training on the same GPU.
PDF arXiv

Activity · 12 weeks

Coming with more sources: repos

Fields · AI · updated 1h ago

Inference & serving

Making models fast and cheap to run: serving systems, kernels, KV-cache, speculative decoding, hardware.

268papers this week−14.1% vs last week

All papers · 1.5k With code · 62

Newest Most relevant

1.5k papers this month

Open problems in Inference & serving

Efficient long-horizon context management3 papers raise it · 3 address it
Long-horizon agents need to retain and reuse information without exceeding finite context windows or degrading performance.
- REMORY: Learning Residual Memory for Context Compactionraises · addresses
- Pack Light, CarryOn: Building Object-Oriented Context for Long-Horizon Agentsraises · addresses
- MemT-1.0: Towards Memory-Native Evolutionraises · addresses
Token-level LLM routing efficiency1 paper raises it · 1 addresses it
Fine-grained token-level routing improves cost-quality trade-offs but strains existing serving systems with desynchronization and complexity.
- TokenRouter: Efficient Serving System for Token-Level LLM Routingraises · addresses
Memory-induced sycophancy in agents1 paper raises it · 1 addresses it
Persistent memories can cause agents to over-align with user beliefs, even when those beliefs are incorrect or outdated.
- Memadapter: Counterfactual Adaptation Against Memory-induced Sycophancyraises · addresses
Quality gap in parallel token generation2 papers raise it · 2 address it
Diffusion language models generate tokens in parallel but lag behind autoregressive models in quality due to uniform computation allocation.
- ALoDLM: Adaptively Looped Diffusion Language Modelsraises · addresses
- Pivot-SD: Efficient Self-Distillation for Masked Diffusion Language Modelsraises · addresses
Sparse attention computation overhead1 paper raises it · 1 addresses it
Sparse attention reduces computation but still requires scoring all keys, limiting efficiency gains.
- DoSA: Dots Sparse Attentionraises · addresses

Named by an AI from these papers' abstracts.