ML-Quant
Papers
6,936 papers featured in the letter since May 2023, each with our one-sentence summary and, where Semantic Scholar knows it, what became of it.
- arXiv
- 2,251
- SSRN
- 2,921
- RePEc
- 928
- Machine learning
- 836
By venue
arXiv
Quantitative-finance and ML-for-finance preprints from arXiv.
- Thinking Fast, Trading Late: System One Decision Models vs the Market
- LLMs Don't Know What They Know: Benchmarking Temporal Self-Knowledge for Financial Backtesting
- The Cryptocurrency Trend Factor Revisited: Predictive Information, Portfolio Construction, and Pricing Claims
2,251 paperssince 24 May 2023
SSRN
Working papers in finance and economics from SSRN.
- The Hidden Sharpe Ratio
- The Financial Impact of Non-Financial Beliefs in Large Language Models
- Does It Actually Run? An Execution-Based Reproducibility Study of Published Machine-Learning Finance Research
2,921 paperssince 24 May 2023
RePEc
Economics working papers from RePEc's NEP field reports.
- Explaining Realized Returns: Roll’s R2 40 Years Later
- Behind the curve: Reassessing inflation risks with a High-dimensional Distributional VAR (HiDVAR)
- Stablecoin Architecture, Digital Dollarization, and the New Triffin Dilemma
928 paperssince 24 May 2023
Machine learning
The general machine-learning papers the letter carried in 2023-25.
- Fairness in Survival Analysis: A Novel Conditional Mutual Information Augmentation Approach
- A comparison of translation performance between DeepL and Supertext
- Revisiting Expected Possession Value in Football: Introducing a Benchmark, U-Net Architecture, and Reward and Risk for Passes
836 paperssince 16 Oct 2023
Most cited
Featured papers with the most citations today (Semantic Scholar).
- 7 Feb 202410,823cites
DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
Advancing Math Reasoning in Language Models: DeepSeekMath7B is a new language model that uses web data and Group Relative Policy Optimization for advanced mathematical reasoning, scoring high on the MATH benchmark.
Machine learningLLMs & Text
- 5 Jun 202410,121cites
Mamba: Linear-Time Sequence Modeling with Selective State Spaces
Sequence Modeling: Mamba, a neural network architecture that doesn't use attention or MLP blocks, provides faster inference and better performance in language, audio, and genomics than Transformers.
Machine learningML & AI Methods
- 16 Oct 20234,163cites
Mistral 7B
Superior Language Model: Mistral 7B v0.1 is a language model with 7 billion parameters that excels in reasoning, mathematics, and code generation, and has a version specifically designed to follow instructions.
Machine learningLLMs & Text
- 9 Oct 20242,560cites
MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark
MMLU-Pro, an improved dataset, expands the Massive Multitask Language Understanding benchmark by adding tougher questions and more choices, serving as a better benchmark to monitor progress in the field.
Machine learningOtherIn Neural Information Processing Systems
- 20 Jun 20242,520cites
Depth Anything V2
Depth Anything V2 is a new model for monocular depth estimation, using synthetic and large-scale pseudo-labeled real images for faster, more accurate results and setting a new evaluation benchmark.
Machine learningOtherIn Neural Information Processing Systems
- 7 Aug 20242,390cites
Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters
The research investigates enhancing Large Language Models' (LLMs) performance using more test-time computation, suggesting a compute-optimal scaling strategy based on prompt difficulty.
Machine learningLLMs & Text
- 22 May 20242,226cites
Octo: An Open-Source Generalist Robot Policy
Octo is a large transformer-based policy for robotic manipulation, trained on a vast dataset, that can be instructed via language or images and adapted to new domains.
Machine learningML & AI MethodsIn Robotics: Science and Systems Conference
- 5 Jun 20242,205cites
Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality
The research identifies a link between state-space models and Transformers in deep learning, leading to the creation of a faster language modeling architecture, Mamba-2.
Machine learningML & AI MethodsIn International Conference on Machine Learning
- 24 Jul 20242,107cites
AWQ: Activation-aware Weight Quantization for On-Device LLM Compression and Acceleration
The study suggests Activation-aware Weight Quantization (AWQ), a hardware-friendly method for quantizing large language models that reduces error and improves performance on various benchmarks.
Machine learningLLMs & TextIn GetMobile: Mobile Computing and Communications
- 12 Dec 20241,838cites
Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling
The paper presents InternVL 2.5, a sophisticated multimodal large language model that performs well on various benchmarks, exceeding 70% on the MMMU benchmark.
Machine learningLLMs & Text
- 16 Oct 20231,806cites
MemGPT: Towards LLMs as Operating Systems
Extended Context in LLMs: MemGPT is a system that manages different memory levels, providing extended context within large language models' limited context windows, enhancing document analysis and multi-session chat performance.
Machine learningLLMs & TextFeatured 2×
- 25 Sep 20241,712cites
Qwen2.5-Coder Technical Report
The report unveils the Qwen2.5-Coder series, an improvement from its predecessor, showcasing remarkable code generation abilities and achieving top-tier performance in various code-related tasks.
Machine learningOtherFeatured 2×
- 5 Feb 20251,613cites
s1: Simple test-time scaling
The research presents a method called budget forcing, which uses a small dataset to achieve test-time scaling and improved reasoning performance in language modeling, particularly in competition math questions.
Machine learningLLMs & TextIn Conference on Empirical Methods in Natural Language Processing
- 22 May 20241,602cites
DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model
MoE Language Model: DeepSeek-V2, a language model with 236B parameters, offers enhanced performance and cost efficiency compared to its predecessor, ranking high among open-source models.
Machine learningLLMs & Text
- 24 Apr 20241,589cites
Mastering Diverse Domains through World Models
Algorithm Mastery: DreamerV3, a universal algorithm, excels in over 150 varied tasks, including diamond collection in Minecraft without human input, expanding the scope of reinforcement learning.
Machine learningML & AI Methods
- 16 Oct 20231,410cites
SelfCheckGPT: Zero-Resource Black-Box Hallucination Detection for Generative Large Language Models
Hallucination Detection for LLMs: The paper presents SelfCheckGPT, a new approach for fact-checking black-box model responses without an external database, proving its superior ability to detect and rank factual and non-factual sentences.
Machine learningLLMs & TextIn Conference on Empirical Methods in Natural Language ProcessingFeatured 2×
- 10 Jul 20241,271cites
SimPO: Simple Preference Optimization with a Reference-Free Reward
Simple Preference Optimization: SimPO improves reinforcement learning from human feedback by using the average log probability of a sequence as the implicit reward, enhancing training stability and computational efficiency.
Machine learningML & AI MethodsIn Neural Information Processing Systems
- 12 Jun 20241,269cites
Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction
The article discusses Visual AutoRegressive modeling (VAR), a new image learning method that outperforms diffusion transformers in terms of speed, image quality, and scalability.
Machine learningML & AI MethodsIn Neural Information Processing Systems
- 8 May 20241,079cites
A Simple and Effective Pruning Approach for Large Language Models
Wanda, a new method, efficiently prunes weights in Large Language Models without retraining, offering a more efficient approach to inducing sparsity in pretrained models.
Machine learningLLMs & TextIn International Conference on Learning Representations
- 12 Jun 20241,066cites
Simple and Effective Masked Diffusion Language Models
The performance of diffusion models in language modeling has been enhanced by using an effective training recipe and a simplified objective, setting a new standard among diffusion models.
Machine learningLLMs & TextIn Neural Information Processing Systems
- 5 Feb 20251,065cites
TÜLU 3: Pushing Frontiers in Open Language Model Post-Training
Open Language Model Post-Training: The Tulu 3 model, a top-tier post-trained language model, is introduced, outperforming other models and providing a detailed guide for its use and adaptation.
Machine learningLLMs & Text
- 3 Jan 20241,032cites
PixArt-α: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis
PIXART-$\alpha$, a Transformer-based text-to-image model, generates high-quality images at a low cost, reducing CO2 emissions and offering a cost-effective solution for the AIGC community.
Machine learningML & AI MethodsIn International Conference on Learning Representations
- 9 Jan 2024982cites
TinyLlama: An Open-Source Small Language Model
Small Open-Source Language Model: The article presents TinyLlama, a compact 1.1B language model that performs remarkably well in various tasks despite its small size, having been pretrained on around 1 trillion tokens.
Machine learningLLMs & Text
- 12 Dec 2024873cites
Training Large Language Models to Reason in a Continuous Latent Space
The article presents Coconut, a new approach that uses the last hidden state of large language models for reasoning in an unrestricted latent space, proving its effectiveness in enhancing the LLM on multiple reasoning tasks.
Machine learningLLMs & Text
- 10 Apr 2024830cites
KVQuant: Towards 10 Million Context Length LLM Inference with KV Cache Quantization
KVQuant, a new method for quantizing cached KV activations in large language models, has been developed, allowing the LLaMA-7B model to be served on an 8-GPU system with minimal degradation.
Machine learningLLMs & TextIn Neural Information Processing Systems
Latest featured
- 6 Nov 20263fanfare
Investing in Data Capital
Measuring data investment from earnings calls, the research shows data capital's share of investment reached about one-eighth in 2025 and commands a market premium of 12 dollars per dollar invested.
- 6 Nov 20263fanfare
The Illusion of Informed Congressional Trading: Evidence from Committee Networks and Hearings
Using machine learning to map firm disclosures to congressional committee jurisdictions, finds no evidence that members earn abnormal returns from legislative access or committee hearings.
- 6 Nov 20263fanfare
The Corporate Bond ETF Creation Puzzle
Finds that authorized participants create corporate bond ETF shares at discounts to offload overvalued inventory, causing delivered bonds to underperform by 5.3 basis points.
- 6 Nov 20264fanfare
The Financial Impact of Non-Financial Beliefs in Large Language Models
Following AI adoption, financial analysts revise high-ESG forecasts upward and become more bullish, but forecast accuracy deteriorates and errors rise for top-ESG firms.
SSRNLLMs & Text
- 6 Nov 20264fanfare
The Hidden Sharpe Ratio
The study estimates maximum Sharpe ratios in large cross-sections without constructing the efficient portfolio, finding annualized maximal Sharpe of 7.4 in U.S. equities versus benchmark estimates of 1.0 to 4.0.
- 6 Nov 20262fanfare
Quantitative Tightening in Segmented Interbank Markets
Models how central banks must retain more reserves for rate control when interbank trading cannot redistribute reserves across heterogeneous banks, even if aggregate demand is sufficient.
- 6 Nov 20262fanfare
AP-PEN: A Hard-Constrained, Physics-Enforced Network for Convenience-Yield Inversion from Commodity Futures
A neural network that analytically enforces the Gibson-Schwartz PDE recovers commodity convenience yield from futures and avoided implausible estimates during the April 2020 WTI dislocation.
- 6 Nov 20264fanfare
Does It Actually Run? An Execution-Based Reproducibility Study of Published Machine-Learning Finance Research
Testing 459 quantitative-finance machine-learning papers' released code, the study finds that execution success rates fall steeply due to environment and dependency issues, not link rot alone.
SSRNML & AI Methods
- 6 Nov 20262fanfare
Is Retail Crypto Trading Better Than Gambling? Memecoins, Perpetual Futures and Prediction Markets Against Casino Odds
Memecoin traders get back 0.754 per unit staked, worse than any casino game, while perpetual futures and prediction markets match casino returns.
SSRNCrypto & DeFi
- 6 Nov 20263fanfare
Not All Beats Are Equal: Anchoring on the Nominal Earnings Surprise
Shows that stock returns respond more to the same percentage surprise when reported earnings are larger in cents, consistent with anchoring on nominal rather than percentage earnings beats.
- 6 Nov 20263fanfare
Shock Transmission Through Bank-Nonbank Linkages: Evidence from Private Credit
Private credit funds that borrow from banks transmit funding shocks to firm lending; following SVB's collapse, exposed banks increased credit-line commitments while cutting direct lending to nonfinancial firms.
- 6 Nov 20263fanfare
Buy the Rumor, Sell the News: When is News Priced in?
Analyzing 1.68 million stock-day events from 4.57 million articles, the study finds that directional price moves concentrate before and at publication day, with little incremental return later.
- 6 Nov 20263fanfare
Monetary Policy Expectations of Sophisticated Investors
Fine-tuned LLM analysis of hedge fund letters reveals policy expectations that predict yield-curve moves beyond market rates and correlate with fund returns.
- 6 Nov 20263fanfare
Green Tastes, Beliefs, and Portfolios: Evidence from Synthetic Investors
Using language models as synthetic investors reveals institutional portfolios tilt greener despite weaker green tastes, and beliefs about green returns turn pessimistic after 2020.
SSRNML & AI Methods
- 6 Nov 20262fanfare
Contract-level Informational Efficiency in Prediction Markets: Measurable Ex Post, Not Predictable Ex Ante
Studying 244 resolved Kalshi markets, the research finds that efficiency is measurable ex post but not predictable ex ante, and naive trading strategies fail against market benchmarks.
- 6 Nov 20263fanfare
Commodity Prices and Sovereign Risk
Research shows trade exposure determines whether commodity-price rises reduce or increase sovereign risk, with exporters gaining from higher prices and importers facing reinforced costs.
- 6 Nov 20263fanfare
Short Squeezes in the Corporate Bond Market
The study identifies two types of squeezes in corporate bonds that degrade price discovery and returns, with lender squeezes predicting negative short-horizon returns and market squeezes predicting positive ones.
- 6 Nov 20264fanfare
How Much Data Does an Information Coefficient Need? The Sampling Distribution of a Cross-Sectional IC, and Why Breadth Does Not Buy Precision
Cross-sectional IC volatility has a floor from factor risk that does not shrink with breadth, capping effective sample size at roughly fifty observations.
- 6 Nov 20263fanfare
Risk-Budgeted Mean-Variance Portfolios
A convex optimization method combines risk budgeting with expected returns, delivering higher average returns than risk parity with lower turnover and concentration than traditional mean-variance portfolios.
- 6 Nov 20263fanfare
What Is (Not) in ABS-EE Data?
Analysis of 19.8 million securitized auto loans reveals that reported data systematically understate risk through selection bias, carry-over reporting gaps, and priced-at-origination hidden credit differences.
- 6 Nov 20263fanfare
Monetary Policy, Expected Dividend Growth, and the Equity Risk Premium
Using equity and risk-free rate term structures, monetary policy shocks affect both expected dividend growth and equity risk premium, with effects beyond duration matching.
- 6 Nov 20263fanfare
Exchange Rate Disconnect with Disagreeing Arbitrageurs
Models how wealth redistribution among arbitrageurs with different beliefs amplifies exchange rate volatility and generates interest-parity wedges, accounting for 62 percent of observed variance.
- 6 Nov 20262fanfare
Scoring Informed Trading in Prediction Markets: A Likelihood Ratio Framework
A likelihood-ratio framework ranks wallets by their information content in binary prediction markets, separating ranking from multiple testing and validating the method through simulation.
- 6 Nov 20262fanfare
Forecasting Stablecoin Depegs: Out-of-Sample Evidence from 219 USD Stablecoins
A gradient-boosted classifier trained on 219 USD-pegged stablecoins predicts which coins will lose their peg within one week, achieving 0.59 average precision on test data.
SSRNCrypto & DeFi
- 6 Nov 20263fanfare
Time-Series Momentum in the ETF Era: Decay of the Long-Short Premium and the Case for Long-Only, Volatility-Targeted Trend Allocation
The classic TSMOM long-short premium has decayed to a zero net Sharpe, but a long-only volatility-targeted version across multiple horizons earns 0.58 with strong crisis protection.
- 6 Nov 20263fanfare
The Hidden Cost of Tokenization: Blockchain Adoption and Native Token Price Discovery
NFT airdrops on Ethereum raise block fees and failure rates, shifting price discovery off-chain and impeding native-token information incorporation critical for proof-of-stake security.
SSRNCrypto & DeFi
- 6 Nov 20262fanfare
The Shape of a Public-Data Volatility Surface: Term-Structure Carry, and the Limits of Measuring Skew
Tests term-structure and skew trading signals in equity volatility from free public data, showing term-structure carry lifts Sharpe ratio to 2.94 but skew lacks sufficient historical data.
- 6 Nov 20263fanfare
The flow, not the headline: point-in-time LLM sentiment and the cross-section of cryptocurrency returns *
Headlines scored by LLM within minutes predict cross-sectional cryptocurrency returns with a daily long-short Sharpe of 2.21 until the signal decays.
SSRNCrypto & DeFi
- 6 Nov 20263fanfare
Who Keeps Your Debt Alive? Creditor Capacity and Corporate Debt Renewal
A 2021 life-insurer capital rule change shows that when incumbent creditors' capacity shrinks, corporate debt issuance falls and firms struggle to refinance maturing debt.
- 6 Nov 20263fanfare
Do AI Models Remember Indian Company Financials? Evidence on Look-Ahead Bias by Firm Size
Large language models recall Indian firm financials from within their training data with accuracy that declines sharply for smaller firms, creating serious look-ahead bias in LLM-based backtests.
SSRNML & AI Methods
- 6 Nov 20263fanfare
Improving inflation forecasting by modeling expectation de-anchoring in low-inflation periods
A model allowing expectations to shift to a low-inflation equilibrium cuts forecast errors roughly 50 percent versus central bank projections for euro area core inflation in the 2010s.
- 6 Nov 20264fanfare
Behind the curve: Reassessing inflation risks with a High-dimensional Distributional VAR (HiDVAR)
HiDVAR, a data-rich distributional framework, improves inflation tail-risk assessment by 28% in CRPS relative to standard benchmarks and performs well in real-time evaluation.
- 6 Nov 20263fanfare
The market's mirror: Revealing investor disagreement with LLMS
Using LLM personas, disagreement over news headlines predicts trading volume, with income and education disagreement most predictive and fundamentals generating least disagreement.
RePEcML & AI Methods
- 6 Nov 20262fanfare
Timing matters: Seasonal anomalies in interest rate pass-through
The research finds that euro-area banks transmit interest rate changes more slowly to lending rates in the fourth quarter during tightening, consistent with year-end volume targets.
- 6 Nov 20263fanfare
Affective Sensitivity and Inflation Expectations across the Income Distribution
Lower-income households form inflation beliefs two and a half times more sensitive to their emotional state than higher-income households, affecting expectations substantially independent of fundamentals.
- 6 Nov 20263fanfare
Central Bank Independence and Sovereign Borrowing
Tighter statutory constraints on central bank lending to government reduce real borrowing costs in developing economies while raising debt-to-GDP, with effects strongest where credibility gains matter most.
- 6 Nov 20262fanfare
Financial Conditions and U.S. Macro Tail Risk
Adding financial conditions to macro forecasts raises recession tail probabilities and lowers growth quantiles, though out-of-sample forecast improvement is inconsistent.
- 6 Nov 20263fanfare
When Cash Flows Turn Negative: Liquidity-Driven Selling by Pension Funds
U.S. pension funds facing negative cash flows sell equities as primary liquidity source rather than rebalancing, absorbing $0.67 per dollar of shock through equity sales.
- 6 Nov 20263fanfare
The Impact of Central Bank Stock Purchases: Evidence from Discontinuities in Policy Rules
Using policy-rule discontinuities, the study finds BOJ stock purchases raise long-term rates but lift prices only after yield curve control pegged rates at zero.
- 6 Nov 20264fanfare
Explaining Realized Returns: Roll’s R2 40 Years Later
Wall Street Journal market summaries explain 30-40% of daily and monthly returns out of sample when filtered for stated reasons, compared to under 10% for macro variables.
RePEcML & AI Methods