arXiv subjects

Maryam Haghifam

Publications and source records attributed to Maryam Haghifam.

4 recordsLinked to original sources

arXiv2026

SeDeM: Selective Decompression of Hidden-State Memories for Long-Context Question Answering

Long-context inference with large language models (LLMs) is costly: self-attention during prefill scales quadratically with sequence length, and the key-value (KV) cache grows with the number of processed tokens. Larger context windows also do not ensure reliable evidence use. Context compression reduces this cost, but many soft-compression methods use LLMs as compressors and rely on compact memory tokens both to preserve information and to condition the decoder. We propose SeDeM, a selective decompression framework that decouples compact memory storage from decoder conditioning. SeDeM stores context as compact hidden-state memory blocks, selects query-relevant blocks, and decompresses only the selected blocks for decoder conditioning. Thus, the decoder avoids both full-context processing and direct generation from highly compressed memory slots. On four long-context QA benchmarks, SeDeM achieves higher QA scores than the compression baselines in our main comparison in both 1B and 3B same-backbone settings, and with the 3B backbone exceeds full-context fine-tuning on three datasets. SeDeM also provides favorable quality--efficiency trade-offs, achieving 1.74--2.46$\times$ lower online time-to-first-token and 1.08--1.10$\times$ higher autoregressive decoding throughput relative to ICAE while maintaining strong answer quality.

cs.CL↗

arXiv2026

H$^{2}$MT: Semantic Hierarchy-Aware Hierarchical Memory Transformer

Transformer-based LLMs achieve strong results on many language tasks; however, long inputs remain challenging because context windows are finite, and prefill latency and memory grow rapidly with prompt length. Flat token-stream processing and chunk-based retrieval can therefore spend substantial computation and context budget on text unrelated to the query. Offline-indexed RAG additionally introduces external storage and index management overhead, and typically appends retrieved evidence as raw text, increasing prefill cost and latency. H^{2}MT makes long-context inference structure-aware: it builds a semantic hierarchy offline, computes a memory embedding for each node via bottom-up post-order aggregation, and routes queries coarse-to-fine at inference to prune irrelevant branches early. On LongBench QA (NarrativeQA, HotpotQA, QASPER) and two structured technical-document settings, H MT achieves favorable quality efficiency trade-offs, delivering competitive ROUGE-L and F1 (where applicable) with lower peak GPU memory and time-to-first-token (TTFT) than prompt compression, memory-token methods, and retrieval-augmented generation baselines.

cs.CL↗

arXiv2024

Design and Evaluation of Crowd-sourcing Platforms Based on Users Confidence Judgments

Crowd-sourcing deals with solving problems by assigning them to a large number of non-experts called crowd using their spare time. In these systems, the final answer to the question is determined by summing up the votes obtained from the community. The popularity of using these systems has increased by facilitation of access to community members through mobile phones and the Internet. One of the issues raised in crowd-sourcing is how to choose people and how to collect answers. Usually, the separation of users is done based on their performance in a pre-test. Designing the pre-test for performance calculation is challenging; The pre-test questions should be chosen in a way that they test the characteristics in people related to the main questions. One of the ways to increase the accuracy of crowd-sourcing systems is to pay attention to people's cognitive characteristics and decision-making model to form a crowd and improve the estimation of the accuracy of their answers to questions. People can estimate the correctness of their responses while making a decision. The accuracy of this estimate is determined by a quantity called metacognition ability. Metacoginition is referred to the case where the confidence level is considered along with the answer to increase the accuracy of the solution. In this paper, by both mathematical and experimental analysis, we would answer the following question: Is it possible to improve the performance of the crowd-sourcing system by knowing the metacognition of individuals and recording and using the users' confidence in their answers?

cs.CE↗

arXiv2024

Proteus: Preserving Model Confidentiality during Graph Optimizations

Deep learning (DL) models have revolutionized numerous domains, yet optimizing them for computational efficiency remains a challenging endeavor. Development of new DL models typically involves two parties: the model developers and performance optimizers. The collaboration between the parties often necessitates the model developers exposing the model architecture and computational graph to the optimizers. However, this exposure is undesirable since the model architecture is an important intellectual property, and its innovations require significant investments and expertise. During the exchange, the model is also vulnerable to adversarial attacks via model stealing. This paper presents Proteus, a novel mechanism that enables model optimization by an independent party while preserving the confidentiality of the model architecture. Proteus obfuscates the protected model by partitioning its computational graph into subgraphs and concealing each subgraph within a large pool of generated realistic subgraphs that cannot be easily distinguished from the original. We evaluate Proteus on a range of DNNs, demonstrating its efficacy in preserving confidentiality without compromising performance optimization opportunities. Proteus effectively hides the model as one alternative among up to $10^{32}$ possible model architectures, and is resilient against attacks with a learning-based adversary. We also demonstrate that heuristic based and manual approaches are ineffective in identifying the protected model. To our knowledge, Proteus is the first work that tackles the challenge of model confidentiality during performance optimization. Proteus will be open-sourced for direct use and experimentation, with easy integration with compilers such as ONNXRuntime.

cs.CR↗