arXiv subjects

Runsong Zhao

Publications and source records attributed to Runsong Zhao.

10 recordsLinked to original sources

arXiv2026

HybridThinker: Efficient Chain-of-Thought Reasoning via Compressed Memory and Transient Thought Steps

Extended chain-of-thought (CoT) traces improve LLM reasoning but incur substantial computational and memory costs. While existing CoT compression methods mitigate this by condensing thought steps into compact representations via memory tokens and retaining only these representations at inference time, the loss of fine-grained information makes subsequent steps more error-prone. To alleviate this, we propose \textbf{HybridThinker}, where in addition to preserved these representations, thought steps are also temporarily retained to provide fine-grained details. However, we observe that naively keeping thought steps accessible to subsequent steps \emph{during training} lets the model bypass memory tokens by retrieving information directly from these steps, leaving the model's ability to compress and retrieve information through memory tokens insufficiently trained. We therefore introduce a hybrid training scheme, in which only some thought steps are directly accessible through attention to subsequent steps, while the other thought steps are masked, forcing the model to use memory tokens for compression and retrieval. Across 4 reasoning benchmarks, HybridThinker matches the uncompressed baseline, advancing the state of the art in CoT compression by 5.8 points on average accuracy with similar inference time. Ablation studies confirm that both temporary thought-step retention and the hybrid training scheme contribute to these gains.

cs.CL↗

arXiv2026

Teacher-Guided Policy Optimization for On-Policy Reasoning Distillation under Large Policy Divergence

On-policy distillation (OPD) has become a promising paradigm for reasoning-oriented post-training of large language models (LLMs), especially when combined with reinforcement learning from verifiable rewards (RLVR). Existing OPD methods rely on reverse KL (RKL)-based teacher supervision over trajectories sampled from the student policy. However, we identify a critical limitation: under large teacher--student policy divergence, RL-driven exploration often produces trajectories outside the teacher distribution, resulting in uninformative negative feedback. To address this, we propose Teacher-Guided Policy Optimization (TGPO), an on-policy reasoning distillation method that remains effective under large policy divergence settings. Rather than relying solely on evaluative supervision, TGPO uses teacher to directly guide token level generation conditioning on student-generated contexts; together with RLVR-style trajectory level rewards, TGPO steers exploration toward improved continuations. Experiments on reasoning benchmarks show that TGPO consistently outperforms existing RKL-based OPD methods and remains robust across different teacher models.

cs.LG↗

arXiv2026

MemoSight: Unifying Context Compression and Multi Token Prediction for Reasoning Acceleration

While chain-of-thought (CoT) reasoning enables LLMs to solve challenging reasoning tasks, the linear growth of the KV cache leads to substantial memory and inference overhead. Existing approaches such as context compression and multi-token prediction (MTP) improve efficiency from two complementary directions by compressing historical tokens and generating future tokens in parallel. However, effectively combining them remains challenging due to their different training paradigms and architectural assumptions. In this work, we propose MemoSight (Memory-Foresight-Based Reasoning), a unified framework that integrates context compression and MTP to improve inference efficiency while preserving CoT performance. MemoSight adopts a shared minimalist design based on special tokens and token-specific positional layouts for both compression and parallel prediction. Experiments on four reasoning benchmarks show that, compared to the vanilla SFT baseline, MemoSight reduces KV cache usage by up to 66% and improves inference speed by 56%, while incurring less than a 3% drop in average reasoning accuracy, yielding a better efficiency-accuracy trade-off than existing CoT compression methods.

cs.AI↗

arXiv2026

PoC: Performance-oriented Context Compression for Large Language Models via Performance Prediction

While context compression can mitigate the growing inference costs of Large Language Models (LLMs) by shortening contexts, existing methods that specify a target compression ratio or length suffer from unpredictable performance degradation, hindering their reliable deployment. We introduce a paradigm shift to Performance-oriented Context Compression (PoC), where developers specify an acceptable performance floor instead of a compression ratio. PoC employs a lightweight performance predictor to automatically find the most aggressive compression ratio that satisfies this constraint before steering an off-the-shelf compressor. We design and compare two predictor variants: a simple context-agnostic predictor and a more sophisticated context-aware one that considers the input's inherent compressibility. On both question-answering and summarization benchmarks, the context-aware predictor consistently achieves lower performance prediction error than the context-agnostic predictor, while the resulting context-aware PoC attains a superior overall performance. Our work paves the way for a more reliable, efficient, and performance-aware deployment of context compression for LLMs.

cs.CL↗

arXiv2026

CoMeT: Collaborative Memory Transformer for Efficient Long Context Modeling

The quadratic complexity and indefinitely growing key-value (KV) cache of standard Transformers pose a major barrier to long-context processing. To overcome this, we introduce the Collaborative Memory Transformer (CoMeT), a novel architecture that enables LLMs to handle arbitrarily long sequences with constant memory usage and linear time complexity. Designed as an efficient, plug-in module, CoMeT can be integrated into pre-trained models with only minimal fine-tuning. It operates on sequential data chunks, using a dual-memory system to manage context: a temporary memory on a FIFO queue for recent events, and a global memory with a gated update rule for long-range dependencies. These memories then act as a dynamic soft prompt for the next chunk. To enable efficient fine-tuning on extremely long contexts, we introduce a novel layer-level pipeline parallelism strategy. The effectiveness of our approach is remarkable: a model equipped with CoMeT and fine-tuned on 32k contexts can accurately retrieve a passkey from any position within a 1M token sequence. On the SCROLLS benchmark, CoMeT surpasses other efficient methods and achieves performance comparable to a full-attention baseline on summarization tasks. Its practical effectiveness is further validated on real-world agent and user behavior QA tasks. The code is available at: https://github.com/LivingFutureLab/Comet

cs.LG↗

arXiv2026

Read As Human: Compressing Context via Parallelizable Close Reading and Skimming

Large Language Models (LLMs) demonstrate exceptional capability across diverse tasks. However, their deployment in long-context scenarios is hindered by two challenges: computational inefficiency and redundant information. We propose RAM (Read As HuMan), a context compression framework that adopts an adaptive hybrid reading strategy, to address these challenges. Inspired by human reading behavior (i.e., close reading important content while skimming less relevant content), RAM partitions the context into segments and encodes them with the input query in parallel. High-relevance segments are fully retained (close reading), while low-relevance ones are query-guided compressed into compact summary vectors (skimming). Both explicit textual segments and implicit summary vectors are concatenated and fed into decoder to achieve both superior performance and natural language format interpretability. To refine the decision boundary between close reading and skimming, we further introduce a contrastive learning objective based on positive and negative query-segment pairs. Experiments demonstrate that RAM outperforms existing baselines on multiple question answering and summarization benchmarks across two backbones, while delivering up to a 12x end-to-end speedup on long inputs (average length 16K; maximum length 32K).

cs.CL↗

arXiv2025

MTP-S2UT: Enhancing Speech-to-Speech Translation Quality with Multi-token Prediction

Current direct speech-to-speech translation methods predominantly employ speech tokens as intermediate representations. However, a single speech token is not dense in semantics, so we generally need multiple tokens to express a complete semantic unit. To address this limitation, we introduce multi-token prediction (MTP) loss into speech-to-unit translation (S2UT) models, enabling models to predict multiple subsequent tokens at each position, thereby capturing more complete semantics and enhancing information density per position. Initial MTP implementations apply the loss at the final layer, which improves output representation but initiates information enrichment too late. We hypothesize that advancing the information enrichment process to intermediate layers can achieve earlier and more effective enhancement of hidden representation. Consequently, we propose MTP-S2UT loss, applying MTP loss to hidden representation where CTC loss is computed. Experiments demonstrate that all MTP loss variants consistently improve the quality of S2UT translation, with MTP-S2UT achieving the best performance.

cs.CL↗

arXiv2025

Autoencoding-Free Context Compression for LLMs via Contextual Semantic Anchors

Context compression is an advanced technique that accelerates large language model (LLM) inference by converting long inputs into compact representations. Existing methods primarily rely on autoencoding tasks to train special compression tokens to represent contextual semantics. While autoencoding tasks enable compression tokens to acquire compression capabilities, we remark that such capabilities potentially conflict with actual downstream task requirements, prevent the models from learning the features more beneficial for real-world usage. Based on this observation, we propose Semantic-Anchor Compression (SAC), a novel method that shifts from autoencoding task based compression to an architecture that is equipped with this compression capability \textit{a priori}. Instead of training models to compress contexts through autoencoding tasks, SAC directly selects so-called anchor tokens from the original context and aggregates contextual information into their key-value (KV) representations. To ensure that anchors can effectively collect information, SAC introduces two key designs: (1) anchor embedding, a learnable embedding vector attached to the selected anchor tokens to mark compression carriers and (2) bidirectional attention modification, which enables anchor tokens to integrate information from the entire context. Experimental results show that SAC consistently outperforms existing context compression methods across different compression ratios and model sizes on question-answering and long-context summarization tasks. Our data, model and code have been released at \href{https://github.com/lx-Meteors/SAC}{https://github.com/lx-Meteors/SAC}.

cs.CL↗

arXiv2024

Forgetting Curve: A Reliable Method for Evaluating Memorization Capability for Long-context Models

Numerous recent works target to extend effective context length for language models and various methods, tasks and benchmarks exist to measure model's effective memorization length. However, through thorough investigations, we find limitations for currently existing evaluations on model's memorization capability. We provide an extensive survey for limitations in this work and propose a new method called forgetting curve to measure the memorization capability of long-context models. We show that forgetting curve has the advantage of being robust to the tested corpus and the experimental settings, of not relying on prompts and can be applied to any model size. We apply our forgetting curve to a large variety of models involving both transformer and RNN/SSM based architectures. Our measurement provides empirical evidence for the effectiveness of transformer extension techniques while raises questions for the effective length of RNN/SSM based models. We also examine the difference between our measurement and existing benchmarks as well as popular metrics for various models. Our code and results can be found at https://github.com/1azybug/ForgettingCurve.

cs.CL↗

arXiv2024

Position IDs Matter: An Enhanced Position Layout for Efficient Context Compression in Large Language Models

Using special tokens (e.g., gist, memory, or compressed tokens) to compress context information is a common practice for large language models (LLMs). However, existing approaches often neglect that position encodings inherently induce local inductive biases in models, causing the compression process to ignore holistic contextual dependencies. We propose \textbf{Enhanced Position Layout (EPL)}, a simple yet effective method that improves the context compression capability of LLMs by only adjusting position IDs, the numerical identifiers that specify token positions. EPL minimizes the distance between context tokens and their corresponding special tokens and at the same time maintains the sequence order in position IDs between context tokens, special tokens, and the subsequent tokens. Integrating EPL into our best performing context compression model results in a 1.9 ROUGE-1 F1 improvement on out-of-domain question answering datasets on average. When extended to multimodal scenarios, EPL leads to an average accuracy gain of 2.6 points for vision compression LLMs.

cs.CL↗