Transformers learn to recall with non-orthogonal embeddings in realistic settings.
problem Understanding how transformers store and retrieve knowledge in practical scenarios.
method Analyzing a single-layer transformer with random embeddings trained on a token-retrieval task.
result Explicit formulas for the model's storage capacity reveal a multiplicative dependence on sample size, embedding dimension, and sequence length.
FinReflectKG benchmarks financial QA by linking relevant context from a financial KG, improving model performance and efficiency.
problem Efficiently retrieving and navigating relevant financial information across diverse sources and years.
method A benchmark built on a temporally indexed financial KG, generating QA pairs via pattern-specific prompts and quality control, evaluating retrieval scenarios.
result KG-guided retrieval yields substantial gains in correctness and token utilization, improving model performance by 24%.
In this paper we study the property of phase retrievability by redundant sysems of vectors under perturbations of the frame set. Specifically we show that if a set $\fc$ of m vectors in the complex Hilbert space of dimension n allows for vector reconstruction from magnitudes of its coefficients, then there is a pertu…
GMC benchmark isolates retrieval in Transformers, revealing max-margin alignment.
problem Understanding how Transformers develop match-and-copy behavior on natural data.
method Introducing Gaussian Match-and-Copy (GMC) as a minimalist benchmark.
result Gradient descent drives parameters to diverge while aligning with max-margin separator.
GENRE retrieves entities autoregressively, improving efficiency and accuracy.
problem Retrieving entities from queries efficiently and accurately.
method Autoregressive generation of entity names, reducing memory footprint and improving context encoding.
result Significantly improved performance on entity disambiguation, linking, and retrieval tasks.
Unified framework for sequence models using test-time regression.
problem Designing efficient sequence models with associative memory.
method Formalizing associative recall as regression over input tokens, deriving various sequence models.
result Clarifies the effectiveness of query-key normalization in softmax attention and offers new generalizations.
Transformer attention layers solve single-location regression tasks.
problem Understanding token-wise sparsity and internal linear representations in attention-based models.
method Introduce single-location regression task and a simplified predictor based on self-attention layers.
result Transformer attention layers are asymptotically Bayes optimal and can learn underlying structures effectively.
We consider the large-scale query-document retrieval problem: given a query (e.g., a question), return the set of relevant documents (e.g., paragraphs containing the answer) from a large document corpus. This problem is often solved in two steps. The retrieval phase first reduces the solution space, returning a subset …
This paper examines how different decoding algorithms for LLMs align with various goals.
problem Consistency of decoding algorithms with different goals in LLMs.
method Analysis of greedy, lookahead, random sampling, and temperature-scaled random sampling algorithms.
result Random sampling is consistent with the true probability distribution, but other goals require optimal algorithms for specific probability distributions.
The paper distinguishes between conditional and marginal processes in language models and discusses conditions for usefulness.
problem The conditional nature of language models trained on observed sequences and the need for marginal text-only processes.
method Distinguishing between full conditional language process, marginal text-only process, and model-induced distribution; analyzing assumptions of stationarity and ergodicity.
result The marginal text-only law is useful only when the observed prefix is an approximately sufficient statistic for the latent circumstances relevant to continuation.
LLMs can be tricked into recalling facts based on context clues.
problem Manipulation of LLMs' factual recall through context changes.
method Mathematical exploration of transformers' associative memory properties.
result Transformers use self-attention and value matrix for associative memory.
This study improves sentence embeddings from BERT models.
problem Capturing the underlying meaning of sentences using BERT models.
method Comprehensive review and testing of various sentence embedding extraction and refinement methods.
result Representation-shaping techniques significantly improve sentence embeddings from BERT-based and simple baseline models.
A new method uses a frozen language model to improve sample efficiency in reinforcement learning.
problem Improving sample efficiency in reinforcement learning with partially observable environments.
method FROZEN Hopfield network and HELM (History Embedding Language Model) method.
result HELM achieves new state-of-the-art results on Minigrid and Procgen environments.
T-EMDE bridges the heterogeneity gap between image and text modalities.
problem Finding similarities between image and text modalities with non-related feature spaces.
method Inspired by EMDE, T-EMDE uses sketches for multimodal operations, avoiding self-attention's quadratic complexity.
result T-EMDE achieves state-of-the-art results and reduces model latency.
Transformers can outperform feedforward and recurrent networks due to dynamic sparsity.
problem Understanding when and why Transformers outperform other neural network architectures.
method Analyzing a sequence-to-sequence data generating model with dynamic sparsity, proving sample complexity differences between feedforward, recurrent, and Transformers.
result Transformers can learn dynamic sparsity models with lower sample complexity than feedforward and recurrent networks.
New insights show embedding lengths correlate with semantic properties.
problem Contrastive embedding norms ignore embedding magnitudes but correlate with semantic properties.
method Formal theoretical framework and analysis of optimization dynamics.
result Embedding lengths encode semantic information as a byproduct of training.
Neural sequence generation is typically performed token-by-token and left-to-right. Whenever a token is generated only previously produced tokens are taken into consideration. In contrast, for problems such as sequence classification, bidirectional attention, which takes both past and future tokens into consideration, …
Detects corruption in agentic models during execution.
problem Inconsistent context, retrieval errors, or adversarial inputs corrupt intermediate steps of reasoning chains.
method Analyzes token graphs induced by attention and computes spectral statistics to emit accept/reject signals.
result A single threshold on the high frequency energy ratio optimally detects context inconsistency in agentic models.
KVCOMM optimizes multi-agent LLM systems by reusing KV-caches, reducing redundant processing.
problem Substantial overhead from reprocessing overlapping contexts across multi-agent systems.
method KVCOMM reuses KV-caches and adjusts offsets for shared content using a pool of cached examples (anchors).
result Achieves over 70% reuse rate across diverse multi-agent tasks, up to 7.8x speedup.
ECLIPSE detects AI hallucinations in finance with high accuracy.
problem Hallucinations in AI-generated answers limit safe deployment in finance.
method Combines entropy estimation and perplexity decomposition to measure model evidence use.
result ECLIPSE achieves ROC AUC of 0.89 and average precision of 0.90 on financial QA dataset.
LITE models improve query-document relevance with learnable late interactions.
problem Improving query-document relevance with lower latency and storage.
method Proposes learnable late-interaction models (LITE) that use factorized query and document embeddings followed by a learnable scorer.
result Empirically, LITE outperforms previous late-interaction models in re-ranking tasks.
Paper reduces vocabulary losslessly for language model cooperation.
problem Language models struggle to cooperate with different tokenizations.
method Established a theoretical framework for lossless vocabulary reduction.
result Efficiently converts models with different tokenizations to cooperate with maximal common vocabulary.
This research improves capital efficiency and impermanent loss in cryptocurrency markets using multi-token trading pools.
problem Poor impermanent loss and capital efficiency in automated market makers.
method Analysis and construction of a multi-token token proactive market maker (MPMM).
result MPMM shows better impermanent loss and capital efficiency than comparable market makers.
This study examines whether tokenized assets improve liquidity and finds significant differences across categories.
problem Improving liquidity for real-world assets through tokenization.
method Examined tokenized real-world assets using Ethereum-based data, measuring liquidity through turnover, active addresses, and active-month indicator.
result Gold-backed tokens show more persistent on-chain activity than Treasury and private-credit-related products, but asset value alone does not reliably predict liquidity.
Study examines revenue from scam tokens on Ethereum, revealing key characteristics and market factors.
problem Revenue from scam tokens on Ethereum blockchain.
method Empirical analysis of Uniswap, examining characteristics and market factors.
result Revenue from scam tokens is influenced by market economic factors and community engagement.
Blockchain-based exchanges adopt based on token pair volatility and personal use.
problem Token value loss and arbitrage issues in decentralized exchanges.
method Investigation of Automated Market Makers (AMMs) using transaction-level data.
result AMMs are adopted for high personal use or highly correlated token price movements.
Blockchain fan tokens boost sports fan engagement by 50%.
problem Low fan engagement in sports decisions.
method Analyzed 3,576 fan token polls to identify determinants of participation.
result Fan tokens engage 4,003 participants per poll, 50% of token holders.
Paper introduces a method to assess liquidity risk in meme tokens using entity-linked address analysis.
problem High market volatility and vulnerability to manipulation in meme tokens.
method Multi-dimensional approach integrating fund flow analysis, behavioral similarity, and anomalous transaction detection.
result Significant disparities between apparent and actual liquidity in meme token markets.
DOS improves language model generation by considering inter-token dependencies.
problem Lack of sequence-level information and inter-token dependencies in existing decoding strategies.
method Dependency-Oriented Sampler (DOS) that uses attention matrices to approximate inter-token dependencies.
result DOS consistently achieves superior performance on code generation and mathematical reasoning tasks.
Study predicts success of crypto-tokens on Pump.fun platform.
problem Identify factors affecting the success of new crypto-tokens.
method Build predictive models using bonding curve mechanism and structural/behavioral variables.
result Conditional variables significantly improve the predictive power of token success.
Sparse attention model reduces long-context inference time with exponential accuracy guarantees.
problem Efficiently processing long-context queries in large language models.
method Formalizes attention as a projection onto key vectors, analyzes entropic relaxation, and introduces Vashista Sparse Attention.
result Sparse attention concentrates on a constant-size active face, leading to exponential decay of inactive tokens' mass and linear scaling of active face error.
Proving that next-token prediction makes language models generate coherent long documents.
problem Understanding why language models generate coherent documents despite focusing on next-token prediction.
method Proving the power of next-token prediction in learning longer-range structure using Recurrent Neural Networks (RNN).
result Optimizing next-token prediction in RNNs yields a model that closely approximates the training distribution, even for long-range coherence.
Expands MLM by masking token positions, improving performance and convergence.
problem Improving language model performance and convergence.
method Masking token positions along with [MASK] tokens, using a fully connected classifier stage.
result Shows .3% improvement and 50% faster convergence for BERT Base with position masking.
This paper compares token and equity financing for startups.
problem Understanding differences in return rates between token and equity financing.
method Developed a three-period model to analyze liquidity and return differences.
result Entrepreneurs can achieve higher payoffs by issuing tokens, especially for risk-averse investors with liquidity needs.
QA-Token improves tokenization for noisy data, boosting model performance.
problem Tokenization ignores data quality, limiting model effectiveness on noisy corpora.
method QA-Token combines signal quality with vocabulary construction through bilevel optimization and reinforcement learning.
result QA-Token achieves state-of-the-art performance on genomic and financial datasets.
Minimal token perturbations reveal how Transformer models process information.
problem Understanding information propagation in Transformer models for interpretability.
method Study of minimal token perturbations on embedding space.
result Rare tokens cause larger shifts, and input information mixes deeper.
LLM-as-a-service prices vary arbitrarily due to tokenization multiplicity.
problem Arbitrary price variation in LLM-as-a-service due to multiple tokenizations of the same output.
method Introduce canonical generation to restrict LLMs to unique tokenizations and develop an efficient sampling algorithm.
result Our sampling algorithm for canonical generation solves tokenization multiplicity and maintains comparable performance and runtime to standard sampling.
Traditional centralized energy systems have the disadvantages of difficult management and insufficient incentives. Blockchain is an emerging technology, which can be utilized in energy systems to enhance their management and control. Integrating token economy and blockchain technology, token economic systems in energy …
New insights show stochastic initialization prevents token clustering in deep Transformers.
problem Understanding token dynamics in deep stochastic Transformers.
method Analysis of deep Transformers with random initialization noise, proving convergence to an interacting-particle system on the sphere.
result Initialization noise prevents token clustering, leading to antipodal formations.
The paper analyzes risk spillovers between AI ETFs, AI tokens, and green markets.
problem Risk spillovers among AI ETFs, AI tokens, and green markets.
method R2 decomposition method
result AI ETFs and clean energy act as risk transmitters, while AI tokens and green assets act as receivers.
Recent years have seen remarkable progress of text generation in different contexts, such as the most common setting of generating text from scratch, and the emerging paradigm of retrieval-and-rewriting. Text infilling, which fills missing text portions of a sentence or paragraph, is also of numerous use in real life, …
Study reveals risks of investing in new crypto-tokens in decentralized exchanges.
problem Risks associated with investing in newly created tokens in decentralized exchanges.
method Analysis of financial impact, market dynamics, profitability, and liquidity manipulations.
result Significant market liquidity trapped in honeypots, reducing market efficiency and misleading investors.
Tokenized RWAs face liquidity issues despite promising markets.
problem Low trading volumes and limited investor participation in tokenized assets.
method Empirical analysis of tokenized real estate, private credit, and treasury funds.
result Most tokenized assets exhibit low transfer activity and limited secondary trading.
Fan tokens surged before World Cup matches, but declined during them, revealing cognitive biases.
problem Analyzing the impact of FIFA World Cup matches on fan tokens.
method Event study and intraday analysis of blockchain-based fan tokens.
result Fan tokens experienced a surge in returns six months before the World Cup, followed by a decline during the matches, revealing asymmetries in performance.
We empirically verify that the market capitalisations of coins and tokens in the cryptocurrency universe follow power-law distributions with significantly different values, with the tail exponent falling between 0.5 and 0.7 for coins, and between 1.0 and 1.3 for tokens. We provide a rationale for this, based on a simpl…
In distributed function computation, each node has an initial value and the goal is to compute a function of these values in a distributed manner. In this paper, we propose a novel token-based approach to compute a wide class of target functions to which we refer as "Token-based function Computation with Memory" (TCM) …
Decoding strategies often exclude human-like tokens, creating a detectable gap in generated text.
problem Decoding strategies exclude contextually appropriate but statistically rare tokens, creating a detectable gap in generated text.
method Analysis of 1.8 million texts across 8 language models, 5 decoding strategies, and 53 hyperparameter configurations.
result 8-18% of human-selected tokens fall outside typical truncation boundaries, indicating a detectable gap.
Hydra boosts efficiency for long-context reasoning in resource-constrained settings.
problem Quadratic complexity of transformers limits long-context reasoning in resource-constrained systems.
method Hydra uses a modular architecture with adaptive routing between sparse global attention, mixture-of-experts, and dual memories.
result Hydra achieves significant throughput and accuracy improvements for long-context reasoning.