Pith. sign in

REVIEW 3 major objections 7 minor 1 cited by

Advancing Event Forecasting through Massive Training of Large Language Models: Challenges, Solutions, and Broader Impacts

T0 review · 3 major / 7 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read Recent advances make it timely to train large language models toward superforecaster-level event forecasting.

desk verdict A genuinely useful, honest position paper that frames the forecasting-training agenda well; its load-bearing transfer assumptions are openly unvalidated, so read it as a research map, not a result. read the letter →

arxiv 2507.19477 v1 pith:JA7K6A3T submitted 2025-07-25 cs.LG cs.AIcs.CL

classification cs.LGcs.AIcs.CL
keywords eventforecastinglargelanguagemodelspredictionmarketssuperforecaster-levelBayesiannetworkstraininglabelsreinforcementlearningknowledgecutoff
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This position paper argues that recent advances in large language models have made large-scale training of superforecaster-level event forecasting systems both timely and feasible. It identifies three obstacles that make event forecasting training unlike other tasks: noisy and sparse outcomes, the knowledge cut-off that devalues pre-cutoff data, and a simple reward structure that rewards overconfident guessing, and it proposes mitigations for each. For label choice, it introduces a hypothetical event Bayesian network showing that intermediate market predictions can be a better training signal than question-time market predictions or final outcomes in certain noise regimes. It also argues for aggressively expanding training data from prediction markets, public databases, and web crawling. If the argument is right, forecasting AI could move beyond matching expert human forecasters to answering questions markets do not cover and powering future simulation and agent reasoning.

What carries the argument

The key analytic object is a hypothetical event Bayesian network with hidden probability $P_{\mathrm{hidden}} = (1-\pi)\alpha + \pi\beta$, where $\alpha$ and $\beta$ are the final success probabilities conditional on negative and positive intermediate states and $\pi$ is the probability of the positive intermediate state. Market predictions $m_0$ and $m_1$ and outcome $o$ are treated as noisy observations of this structure, and simulation shows a bias-variance trade-off: $m_0$ has lowest variance but highest bias, $o$ has highest variance but lowest bias, and $m_1$ can win in the middle regime. This machinery justifies label-selection strategies in Section 4.1 and extends to using intermediate model predictions $q_1$ as training signals for questions without clear resolution conditions.

What would settle it

Use the same base model and the same set of unresolved market questions, train with $m_0$ labels on one run and $m_1$ labels on another, and compare Brier scores on a held-out dynamic benchmark of questions resolved after training; if $m_1$-trained models do not beat $m_0$-trained models in the regime the paper's simulation identifies, the central sub-claim fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that the time is ripe for research on large-scale training of superforecaster-level event forecasting LLMs. The supporting sub-claim, developed through a hypothetical event Bayesian network and simulation, is that intermediate market predictions $m_1$ can be a superior training label to question-time market predictions $m_0$ and final outcomes $o$ under certain conditions—specifically when there is significant uncertainty about the transition between the initial and intermediate states and a moderate number of training observations. The paper also claims that three mitigations—using poorly-recalled historical events and counterfactual events, and adding auxiliary reward signals—can overcome the knowledge cut-off and simple reward structure problems, and that combining market, public, and crawling datasets can expand training scale from roughly ten thousand instances to more than one hundred thousand.

Load-bearing premise

Training on historical events that the model cannot recall, and on counterfactual events, transfers to improved performance on general event forecasting tasks; the paper itself flags this as an open empirical question.

Editorial extensions

If this is right

  • Label choice becomes a tunable training hyperparameter: use outcomes where similar events are plentiful, market predictions where they are sparse, and intermediate market predictions in between.
  • Pre-cutoff training data can be made usable through poorly-recalled comparative events and counterfactual scenarios, unlocking a much larger pool of historical training instances.
  • Auxiliary rewards from reasoning evaluation and subquestions could keep models from collapsing to overconfident 0% or 100% predictions during RL training.
  • Aggressive use of market, public, and crawling datasets enables both larger-scale training and faster dynamic benchmark evaluation, accelerating development of proprietary forecasting systems.
  • If training succeeds, forecasting AI could answer personalized and ambiguous questions that prediction markets cannot, and its probabilistic reasoning could be integrated into general agents and AI scientists.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial: The paper's simulation suggests a direct head-to-head experiment—training identical models with $m_0$, $m_1$, and $o$ labels on the same data—that would validate or refute the central sub-claim before committing to a large-scale training run.
  • Editorial: If transfer from counterfactual or pre-cutoff training fails, the data-scale agenda in Section 5 still stands, because market and public datasets alone can plausibly reach the 100,000-instance scale; the paper somewhat over-couples these two arguments.
  • Editorial: The self-fulfilling prophecy risk the paper describes could also corrupt its own training labels once AI forecasts influence markets and public outcomes, creating a distribution shift that the proposed data pipeline does not address.
  • Editorial: The claim that AI already beats average individual forecasters suggests that consumer-facing personalized predictions may arrive before superforecaster-level AI, making reliability communication rather than raw accuracy the near-term bottleneck.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 7 minor

Summary. This position paper argues that recent developments in LLM-based event forecasting — steadily improving results on dynamic benchmarks (Karger et al., 2025), outcome-based reinforcement learning results (Turtel et al., 2025a;b), and the emergence of reasoning and Deep Research-style models — indicate that 'the time is ripe for research on large-scale training of superforecaster-level event forecasting LLMs.' The paper identifies three training difficulties specific to event forecasting: noisiness and sparsity, the knowledge cut-off, and the simple reward structure. It then proposes mitigations: a hypothetical-event Bayesian-network framework for choosing training labels (§4.1.2 and Appendix A), training on poorly recalled and counterfactual pre-cutoff events (§4.2), and auxiliary reward signals from reasoning evaluation and subquestions (§4.3). For data, it advocates aggressive use of market, public, and crawled datasets, projecting a scale-up from roughly 10^4 to over 10^5 training instances (§5). The final sections survey societal impacts and risks. The main technical artifact is a simulation in Appendix A showing that intermediate market predictions m1 can, under certain noise conditions, serve as a better training label for the hidden question-time probability Phidden than either question-time market predictions m0 or final outcomes o.

Significance. The paper has genuine strengths: it is unusually honest about the field's evaluation pitfalls, it explicitly names its own key open question (§4.2.2), and the appendix simulation is specified precisely enough to be checked and reproduced. The framing of the knowledge cut-off as the central data-scale bottleneck is insightful, and the market/public/crawling dataset taxonomy gives the agenda a concrete and actionable shape. If the underlying transfer assumptions hold, the program would be high-impact and would close a real gap between current work and superforecaster-level performance. However, both load-bearing premises — transfer from poorly recalled and counterfactual historical training to general event forecasting, and the empirical realizability of the noise regime in which m1 dominates — are unvalidated, and the first is conceded by the paper itself. The stress-test concern about §4.2.2 therefore lands: the scale-up argument in §5 depends on exactly the assumption that the paper flags as open.

major comments (3)
  1. [Abstract; §3; §4.2.2; §5] The Abstract and §3 state the readiness claim ('the time is ripe for research on large-scale training') unconditionally, yet the agenda's main new lever — the projected scale-up from ~10^4 to ≥10^5 training instances in §5 — depends on pre-cutoff training data (§4.2) and on intermediate-prediction labels (§4.1.5 and §5.4), and §4.2.2 explicitly concedes that 'it remains an open empirical question whether training on such poorly-recalled historical events improves performance on general event forecasting tasks.' If that transfer fails, the scale-up argument loses its main new lever and reduces to the already-explored regime of post-cutoff data with outcome/market labels. I recommend that the Abstract and Conclusion either carry this conditionality explicitly or commit to a concrete falsifiable first milestone, such as a pre-registered study that trains on poorly recalled historical event pairs and evaluates on a dynamic benchmark (ForecastBench or the Metaculus AI Benchmarking series); the readiness claim should be tied to such a test rather than to extrapolation from the positive trends in §3.
  2. [Appendix A; Fig. 2; §4.1.2; §4.1.5; §5.4] In Appendix A (Fig. 2), the superiority of m1 over m0 and o is a property of the assumed generative model, not an empirical discovery: m1 is sampled from (1−t)·α̂ + t·β̂ conditional on the realized intermediate state t, and the six noise parameters (σ²_biased and σ²_unbiased for α, β, and π) are free, so the m1-superiority regime in the first panel is produced by construction. What is missing is the bridge to the real training setting. §4.1.2's closing recommendation maps the simulation's N (the number of i.i.d. trials whose estimates are averaged) onto the training-data N (the number of similar events in a training corpus), but an LLM trained on N individual labels is not an estimator that averages those labels; moreover, the averaged m1 is unbiased for Phidden in the simulation precisely because m1 conditions on the realized S1, and the analogous noise structure in real markets is unknown. I ask the authors to add, or explicitly mark as a prerequisite for the §4.1.5 and §5.4 proposals, an empirical label-quality analysis on resolved prediction-market questions (e.g., the empirical Brier scores of m0, m1, and o as targets for question-time predictions, stratified by market depth), so that the m1-based proposals are presented as conditional on that evidence.
  3. [§4.2.3; §4.2.2; §3] §4.2.3's counterfactual-event pipeline is supported only by analogy to counterfactual QA training: Neeman et al. (2023) train models to suppress parametric knowledge in favor of retrieved context for answering factual counterfactual questions, whereas the proposed pipeline requires the model to produce calibrated probability estimates from fictional news documents during training and to transfer that probabilistic reasoning to real questions at inference. The paper does not address why probability-calibration behavior learned under generated counterfactual outcomes would transfer to a real distribution in which the model's prior knowledge is partly valid. In addition, the proposed divergence-date selection criterion of §4.2.3 (query the model's probability at past time points and sample dates where uncertainty is 'appropriately high') risks encoding the model's own miscalibration into the training distribution. I recommend specifying a validation loop — for example, measuring ECE and Brier score on a held-out post-cutoff set after counterfactual training — or relabeling this proposal as an explicitly speculative direction rather than one of the paper's solutions.
minor comments (7)
  1. [Fig. 2 caption] The caption refers to the 'average of m0, m1, and m2', but the model in §4.1.2 and Appendix A defines only m0, m1, and o; 'm2' should be 'o'.
  2. [§2.4] The RetroSearch technique is attributed to 'Bosse et al. (2025)', but the cited Bosse et al. paper is the Deep Research Bench study; please verify the attribution, since the primary source for RetroSearch appears to be Wildman et al. (2025), cited in the same sentence.
  3. [§4.1.5] Notation is inconsistent within the section: 'm₁' appears alongside 'm1', 'q1', and 'm0'; the subscript style should be unified.
  4. [Table 1; §4.1] The SpaceX example in Table 1 (referenced to Polymarket 2023) is a different question from the orbital-flight-by-June-2024 example used in §4.1 and §4.1.2; the running example should be aligned throughout the paper.
  5. [§4.2.3] There is a spacing typo in 'counterfactualoutcomesandusethemasiftheyweresearchresults'; the paragraph should receive a copyedit pass.
  6. [§3] The claim that AI 'has likely already significantly surpassed average individual performance' would benefit from an explicit statement of the comparison baseline, since the numbers reported just before (median general-public prediction 0.121 vs. GPT-4o 0.133 and Claude-3.5-Sonnet 0.122) show the crowd median roughly matching the best models; clarifying that the comparison is against average individuals, not the crowd median, would avoid confusion.
  7. [§4.1.2 and §4.1.3] There is a tension between the conclusion that outcomes are the best label when the number of similar events is large (§4.1.2) and the warning that outcome-based training pushes models toward extreme 0/1 predictions and thereby hinders search and reasoning (§4.1.3); the paper should spell out how the label-quality analysis relates to the reward-structure critique.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: agenda grounding is external, and the m1-label analysis is an explicitly hypothetical model rather than a fitted prediction.

full rationale

This position paper's central claim is that recent external evidence makes large-scale training of event-forecasting LLMs timely. That evidence comes from independent sources (Karger et al. 2025; Turtel et al. 2025a,b; Metaculus reports; Deep Research deployments), not from the present authors' prior results, so no load-bearing self-citation chain is present. The only candidate for a by-construction result is Appendix A's demonstration that intermediate market prediction m1 can outperform m0 and o under high pi-noise and moderate N; that ordering is, indeed, built into the sampling equations, since m1 is drawn conditional on the realized intermediate state t using noisy alpha-hat and beta-hat, while m0 also carries the pi-hat noise. However, the paper explicitly frames the Bayesian network as hypothetical ('we can assume they estimate values based on noisy observation'; 'We provide detailed assumptions and simulation results demonstrating these trade-offs') and does not present the simulation as an empirical discovery. It also flags the missing empirical validation ('The statistical validity can be established through empirical data analysis'; 'it remains an open empirical question whether training on such poorly-recalled historical events improves performance on general event forecasting tasks'). A theorem about one's own stated assumptions is not circularity. No fitted parameter is renamed as a prediction, and the paper is self-contained against external benchmarks, so no significant circularity is found.

Assumptions & free parameters 1 free parameters · 6 assumptions · 0 invented entities

The central claim is an agenda, so its ledger is mostly assumptions about how markets, memories, and rewards behave. The only explicit free parameters are the hand-chosen noise terms in the Appendix A simulation, which drive the m1-superiority example. No new physical or computational entities are introduced.

free parameters (1)
  • Biased and unbiased noise terms for alpha, beta, pi in the Appendix A simulation = (0.1,0.1,0.2) and (0.2,0.2,0.4); (0.2,0.2,0.4) and (0.2,0.2,0.4); (0.4,0.4,0.4) and (0.4,0.4,0.4)
    Chosen by hand to illustrate different noise scenarios; the m1-superiority conclusion depends on these values.
assumptions (6)
  • standard math Brier score, logarithmic score, ECE definitions and standard probability calculus
    Used throughout Sections 2 and 4 without proof.
  • domain assumption Market predictions are noisy observations of a well-defined hidden probability Phidden, with noise added in log-probability space
    Appendix A defines m0 and m1 as addNoise(alpha, beta, pi) with specified Gaussian noise; this is an assumption about markets, not derived.
  • ad hoc to paper The two-stage structure (S0 -> S1 -> outcome) with parameters alpha, beta, pi captures event forecasting dynamics
    Section 4.1.2 introduces this toy model; real events have richer dynamics.
  • domain assumption Training on counterfactual events reduces memorization and improves reasoning
    Section 4.2.3 extrapolates from counterfactual QA literature (Neeman et al. 2023) to forecasting; untested.
  • domain assumption Subquestions share causal factors and provide independent reward signals
    Section 4.3.3 proposes subquestion auxiliary rewards without empirical validation.
  • domain assumption Model's intermediate prediction q1 can serve as a training signal
    Section 4.1.5 proposes using q1 when markets are unavailable; validity depends on model accuracy, acknowledged in text.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Advancing Event Forecasting through Massive Training of Large Language Models: Challenges, Solutions, and Broader Impacts." pith.science (2026). https://pith.science/paper/JA7K6A3T

@misc{pith2026250719477,
  author       = {Pith},
  title        = {Pith review of: Advancing Event Forecasting through Massive Training of Large Language Models: Challenges, Solutions, and Broader Impacts},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JA7K6A3T}},
  note         = {Machine review of arXiv:2507.19477}
}
read the original abstract

Many recent papers have studied the development of superforecaster-level event forecasting LLMs. While methodological problems with early studies cast doubt on the use of LLMs for event forecasting, recent studies with improved evaluation methods have shown that state-of-the-art LLMs are gradually reaching superforecaster-level performance, and reinforcement learning has also been reported to improve future forecasting. Additionally, the unprecedented success of recent reasoning models and Deep Research-style models suggests that technology capable of greatly improving forecasting performance has been developed. Therefore, based on these positive recent trends, we argue that the time is ripe for research on large-scale training of superforecaster-level event forecasting LLMs. We discuss two key research directions: training methods and data acquisition. For training, we first introduce three difficulties of LLM-based event forecasting training: noisiness-sparsity, knowledge cut-off, and simple reward structure problems. Then, we present related ideas to mitigate these problems: hypothetical event Bayesian networks, utilizing poorly-recalled and counterfactual events, and auxiliary reward signals. For data, we propose aggressive use of market, public, and crawling datasets to enable large-scale training and evaluation. Finally, we explain how these technical advances could enable AI to provide predictive intelligence to society in broader areas. This position paper presents promising specific paths and considerations for getting closer to superforecaster-level AI technology, aiming to call for researchers' interest in these directions.

Figures

Figures reproduced from arXiv: 2507.19477 by the authors.

Figure 1
Figure 1. A hypothetical event Bayesian network. The state [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. Results of the simulation. The x-axis represents [PITH_FULL_IMAGE:figures/full_fig_p029_2.png] view at source ↗

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Revealed Rationality: Label-Free Evaluation and Regularization from Representation Theorems

    econ.TH 2026-08 conditional novelty 6.0 of 10

    Representation theorems from decision theory yield label-free, exhaustive rationality checks and continuous penalties for LLM behavior.

Reference graph

Works this paper leans on

99 extracted references · 57 canonical work pages · cited by 1 Pith paper

  1. [1]

    Forecastqa: A question answering challenge for event forecasting with temporal text data

    Woojeong Jin, Rahul Khanna, Suji Kim, Dong-Ho Lee, Fred Morstatter, Aram Galstyan, and Xiang Ren. Forecastqa: A question answering challenge for event forecasting with temporal text data. In ACL, 2021. URL https://aclanthology.org/2021.acl-long.357/

  2. [2]

    Forecasting future world events with neural networks

    Andy Zou, Tristan Xiao, Ryan Jia, Joe Kwon, Mantas Mazeika, Richard Li, Dawn Song, Jacob Steinhardt, Owain Evans, and Dan Hendrycks. Forecasting future world events with neural networks. NeurIPS, 2022. URL https://proceedings.neurips.cc/paper_files/paper/2022/file/aec870a6772336c15dac992c16f2e7c9-Paper-Datasets_and_Benchmarks.pdf

  3. [3]

    Approaching human-level forecasting with language models

    Danny Halawi, Fred Zhang, Chen Yueh-Han, and Jacob Steinhardt. Approaching human-level forecasting with language models. In NeurIPS, 2024. URL https://openreview.net/forum?id=FlcdW7NPRY

  4. [4]

    u ttler, Mike Lewis, Wen-tau Yih, Tim Rockt \

    Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich K \"u ttler, Mike Lewis, Wen-tau Yih, Tim Rockt \"a schel, et al. Retrieval-augmented generation for knowledge-intensive nlp tasks. NeurIPS, 2020. URL https://proceedings.neurips.cc/paper/2020/hash/6b493230205f780e1bc26945df7481e5-Abstract.html

  5. [5]

    Superforecasting: The art and science of prediction

    Philip E Tetlock and Dan Gardner. Superforecasting: The art and science of prediction. Random House, 2016. URL https://dgs.swanlibraries.net/Record/a1370496?page=&recordIndex=2&referred=resultIndex&searchId=151262722

  6. [6]

    Forecastbench: A dynamic benchmark of AI forecasting capabilities

    Ezra Karger, Houtan Bastani, Chen Yueh-Han, Zachary Jacobs, Danny Halawi, Fred Zhang, and Philip E Tetlock. Forecastbench: A dynamic benchmark of AI forecasting capabilities. In ICLR, 2025. URL https://openreview.net/forum?id=lfPkGWXLLf

  7. [7]

    Q3 ai benchmarking: Did bots outperform human forecasters?, 2024 a

    Tom Liptay. Q3 ai benchmarking: Did bots outperform human forecasters?, 2024 a . URL https://www.metaculus.com/notebooks/28784/aibq3results/

  8. [8]

    Introducing chatgpt, 2022

    OpenAI. Introducing chatgpt, 2022. URL https://openai.com/index/chatgpt/

Show all 99 references
  1. [9]

    Wisdom of the silicon crowd: LLM ensemble prediction capabilities rival human crowd accuracy

    Philipp Schoenegger, Indre Tuminauskaite, Peter S Park, Rafael Valdece Sousa Bastos, and Philip E Tetlock. Wisdom of the silicon crowd: LLM ensemble prediction capabilities rival human crowd accuracy. Science Advances, 10 0 (45), 2024. URL https://www.science.org/doi/10.1126/s...

  2. [10]

    Reasoning and tools for forecasting

    Elvis Hsieh, Preston Fu, and Jonathan Chen. Reasoning and tools for forecasting. In The 4th Workshop on Mathematical Reasoning and AI at NeurIPS'24, 2024. URL https://openreview.net/forum?id=gdxUJvPng5

  3. [11]

    Superhuman automated forecasting, 2024

    Long Phan, Andrew Zeng, Mantas Mazeika, Adam Khoja, and Dan Hendrycks. Superhuman automated forecasting, 2024. URL https://safe.ai/blog/forecasting

  4. [12]

    The memorization problem: Can we trust llms' economic forecasts? arXiv [q-fin.GN], 2025

    Alejandro Lopez-Lira, Yuehua Tang, and Mingyin Zhu. The memorization problem: Can we trust llms' economic forecasts? arXiv [q-fin.GN], 2025. URL https://arxiv.org/abs/2504.14765

  5. [13]

    Contra papers claiming superhuman AI forecasting, 2024

    Nikos Bosse, Peter Mühlbacher, Lawrence Phillips, and Dan Schwarz. Contra papers claiming superhuman AI forecasting, 2024. URL https://www.lesswrong.com/posts/uGkRcHqatmPkvpGLq/contra-papers-claiming-superhuman-ai-forecasting

  6. [14]

    Pitfalls in evaluating language model forecasters

    Daniel Paleka, Shashwat Goel, Jonas Geiping, and Florian Tramèr. Pitfalls in evaluating language model forecasters. arXiv [cs.LG], 2025 a . URL https://arxiv.org/abs/2506.00723

  7. [15]

    Why humans are still much better than ai at forecasting the future, 2025

    Dylan Matthews. Why humans are still much better than ai at forecasting the future, 2025. URL https://www.vox.com/future-perfect/411742/ai-forecasting-prediction-metaculus-llm

  8. [16]

    Introducing openai o1, 2024

    OpenAI. Introducing openai o1, 2024. URL https://openai.com/o1/

  9. [17]

    Introducing openai o3 and o4-mini, 2025 a

    OpenAI. Introducing openai o3 and o4-mini, 2025 a . URL https://openai.com/index/introducing-o3-and-o4-mini/

  10. [18]

    Q4 ai benchmarking: Bots are closing the gap, 2025

    Molly Hickman. Q4 ai benchmarking: Bots are closing the gap, 2025. URL https://www.metaculus.com/notebooks/35291/q4-ai-benchmarking-results/

  11. [19]

    LLMs can teach themselves to better predict the future

    Benjamin Turtel, Danny Franklin, and Philipp Schoenegger. LLMs can teach themselves to better predict the future. arXiv [cs.CL], 2025 a . URL https://arxiv.org/abs/2502.05253

  12. [20]

    Outcome-based reinforcement learning to predict the future

    Benjamin Turtel, Danny Franklin, Kris Skotheim, Luke Hewitt, and Philipp Schoenegger. Outcome-based reinforcement learning to predict the future. arXiv [cs.LG], 2025 b . URL https://arxiv.org/abs/2505.17989

  13. [21]

    Try deep research and our new experimental model in gemini, your ai assistant, 2024

    Dave Citron. Try deep research and our new experimental model in gemini, your ai assistant, 2024. URL https://blog.google/products/gemini/google-gemini-deep-research/

  14. [22]

    Introducing deep research, 2025 b

    OpenAI. Introducing deep research, 2025 b . URL https://openai.com/index/introducing-deep-research/

  15. [23]

    Introducing claude 4, 2025

    Anthropic. Introducing claude 4, 2025. URL https://www.anthropic.com/news/claude-4

  16. [24]

    Will spacex's 3rd starship go higher than its prior launch before april?, 2023

    Polymarket. Will spacex's 3rd starship go higher than its prior launch before april?, 2023. URL https://polymarket.com/event/will-spacexs-3rd-starship-launch-reach-150km-before-april

  17. [25]

    Can language models use forecasting strategies? arXiv [cs.LG], 2024

    Sarah Pratt, Seth Blumberg, Pietro Kreitlon Carolino, and Meredith Ringel Morris. Can language models use forecasting strategies? arXiv [cs.LG], 2024. URL https://arxiv.org/abs/2406.04446

  18. [26]

    https://www.metaculus.com/notebooks/34747/the-state-of-metaculus/, 2025 a

    Christian Williams. https://www.metaculus.com/notebooks/34747/the-state-of-metaculus/, 2025 a . URL https://www.metaculus.com/notebooks/34747/the-state-of-metaculus/

  19. [27]

    What can we learn from scoring different election forecasts?, 2022

    Jack Chen. What can we learn from scoring different election forecasts?, 2022. URL https://firstsigma.substack.com/p/midterm-elections-forecast-comparison-analysis

  20. [28]

    Online prediction betting markets look ahead after us presidential election triumph: `we're just getting started', 2024

    Callum Jones. Online prediction betting markets look ahead after us presidential election triumph: `we're just getting started', 2024. URL https://www.theguardian.com/us-news/2024/nov/15/online-betting-markets-expansion-after-presidential-election

  21. [29]

    Autocast++: Enhancing world event prediction with zero-shot ranking-based context retrieval

    Qi Yan, Raihan Seraj, Jiawei He, Lili Meng, and Tristan Sylvain. Autocast++: Enhancing world event prediction with zero-shot ranking-based context retrieval. In ICLR, 2024. URL https://openreview.net/forum?id=COYDmKkQH4

  22. [30]

    Mirai: Evaluating llm agents for event forecasting

    Chenchen Ye, Ziniu Hu, Yihe Deng, Zijie Huang, Mingyu Derek Ma, Yanqiao Zhu, and Wei Wang. Mirai: Evaluating llm agents for event forecasting. arXiv [cs.CL], 2024. URL https://arxiv.org/abs/2407.01231

  23. [31]

    Consistency checks for language model forecasters

    Daniel Paleka, Abhimanyu Pallavi Sudhir, Alejandro Alvarez, Vineeth Bhat, Adam Shen, Evan Wang, and Florian Tram \`e r. Consistency checks for language model forecasters. In ICLR, 2025 b . URL https://openreview.net/forum?id=r5IXBlTCGc

  24. [32]

    Bench to the future: A pastcasting benchmark for forecasting agents

    Jack Wildman, Nikos I Bosse, Daniel Hnyk, Peter M \"u hlbacher, Finn Hambly, Jon Evans, Dan Schwarz, Lawrence Phillips, et al. Bench to the future: A pastcasting benchmark for forecasting agents. arXiv [cs.CL], 2025. URL https://arxiv.org/abs/2506.21558

  25. [33]

    Deep research bench: Evaluating ai web research agents

    Nikos I Bosse, Jon Evans, Robert G Gambee, Daniel Hnyk, Peter M \"u hlbacher, Lawrence Phillips, Dan Schwarz, Jack Wildman, et al. Deep research bench: Evaluating ai web research agents. arXiv [cs.AI], 2025. URL https://arxiv.org/abs/2506.06287

  26. [34]

    Openforecast: A large-scale open-ended event forecasting dataset

    Zhen Wang, Xi Zhou, Yating Yang, Bo Ma, Lei Wang, Rui Dong, and Azmat Anwar. Openforecast: A large-scale open-ended event forecasting dataset. In COLING, 2025. URL https://aclanthology.org/2025.coling-main.353/

  27. [35]

    scope sensitive

    Tom Liptay. How "scope sensitive" are o1-preview and gpt4o's forecasts? not very., 2024 b . URL https://www.metaculus.com/notebooks/30889/self-consistency-of-gpt4o-vs-o1-preview/

  28. [36]

    Calibrating large language models with sample consistency

    Qing Lyu, Kumar Shridhar, Chaitanya Malaviya, Li Zhang, Yanai Elazar, Niket Tandon, Marianna Apidianaki, Mrinmaya Sachan, and Chris Callison-Burch. Calibrating large language models with sample consistency. In AAAI, 2025. URL https://ojs.aaai.org/index.php/AAAI/article/view/34120

  29. [37]

    Ai forecasting bots incoming, 2024

    Dan Hendrycks and Mantas Mazeika. Ai forecasting bots incoming, 2024. URL https://www.lesswrong.com/posts/4kuXNhPf9FBwok7tK/ai-forecasting-bots-incoming

  30. [38]

    Comparing two forecasters in an ideal world, 2023 a

    Nikos Bosse. Comparing two forecasters in an ideal world, 2023 a . URL https://www.lesswrong.com/posts/uGkRcHqatmPkvpGLq/contra-papers-claiming-superhuman-ai-forecasting

  31. [39]

    Chatbot arena: An open platform for evaluating llms by human preference

    Wei-Lin Chiang, Lianmin Zheng, Ying Sheng, Anastasios Nikolas Angelopoulos, Tianle Li, Dacheng Li, Banghua Zhu, Hao Zhang, Michael Jordan, Joseph E Gonzalez, et al. Chatbot arena: An open platform for evaluating llms by human preference. In ICML, 2024. URL https://openreview.n...

  32. [40]

    Q1 ai benchmark results: Pro forecasters crush bots, 2025

    Ben Wilson and John Bash. Q1 ai benchmark results: Pro forecasters crush bots, 2025. URL https://www.metaculus.com/notebooks/38673/q1-ai-benchmarking-results/

  33. [41]

    https://www.metaculus.com/notebooks/36949/ai-forecasting-benchmark-q2-tournament-starts-april-21/, 2025 b

    Christian Williams. https://www.metaculus.com/notebooks/36949/ai-forecasting-benchmark-q2-tournament-starts-april-21/, 2025 b . URL https://www.metaculus.com/notebooks/36949/ai-forecasting-benchmark-q2-tournament-starts-april-21/

  34. [42]

    T " ulu 3: Pushing frontiers in open language model post-training

    Nathan Lambert, Jacob Morrison, Valentina Pyatkin, Shengyi Huang, Hamish Ivison, Faeze Brahman, Lester James V Miranda, Alisa Liu, Nouha Dziri, Shane Lyu, et al. T " ulu 3: Pushing frontiers in open language model post-training. arXiv [cs.CL], 2024. URL https://arxiv.org/abs/2...

  35. [43]

    Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning

    Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv [cs.CL], 2025. URL https://arxiv.org/abs/2501.12948

  36. [44]

    Futuresearch github, 2025

    Futuresearch. Futuresearch github, 2025. URL https://github.com/futuresearch

  37. [45]

    Can llm generate regression tests for software commits? arXiv [cs.SE], 2025

    Jing Liu, Seongmin Lee, Eleonora Losiouk, and Marcel B \"o hme. Can llm generate regression tests for software commits? arXiv [cs.SE], 2025. URL https://arxiv.org/abs/2501.11086

  38. [46]

    What uncertainties do we need in bayesian deep learning for computer vision? NIPS, 2017

    Alex Kendall and Yarin Gal. What uncertainties do we need in bayesian deep learning for computer vision? NIPS, 2017. URL https://proceedings.neurips.cc/paper/2017/hash/2650d6089a6d640c5e85b2b88265dc2b-Abstract.html

  39. [47]

    More is probably more — forecasting accuracy and number of forecasters on metaculus, 2023 b

    Nikos Bosse. More is probably more — forecasting accuracy and number of forecasters on metaculus, 2023 b . URL https://www.metaculus.com/notebooks/14951/14951/more-is-probably-more-forecasting-accuracy-and-number-of-forecasters-on-metaculus/

  40. [48]

    Predicting empirical ai research outcomes with language models

    Jiaxin Wen, Chenglei Si, Yueh han Chen, He He, and Shi Feng. Predicting empirical ai research outcomes with language models. arXiv [cs.AI], 2025. URL https://arxiv.org/abs/2506.00794

  41. [49]

    Retrieval-guided counterfactual generation for QA

    Bhargavi Paranjape, Matthew Lamm, and Ian Tenney. Retrieval-guided counterfactual generation for QA . In ACL, 2022. URL https://aclanthology.org/2022.acl-long.117/

  42. [50]

    Disentqa: Disentangling parametric and contextual knowledge with counterfactual question answering

    Ella Neeman, Roee Aharoni, Or Honnovich, Leshem Choshen, Idan Szpektor, and Omri Abend. Disentqa: Disentangling parametric and contextual knowledge with counterfactual question answering. In ACL, 2023. URL https://aclanthology.org/2023.acl-long.559/

  43. [51]

    Reinforcement learning with unsupervised auxiliary tasks

    Max Jaderberg, Volodymyr Mnih, Wojciech Marian Czarnecki, Tom Schaul, Joel Z Leibo, David Silver, and Koray Kavukcuoglu. Reinforcement learning with unsupervised auxiliary tasks. In ICLR, 2017. URL https://openreview.net/forum?id=SJ6yPD5xg

  44. [52]

    The biggen bench: A principled benchmark for fine-grained evaluation of language models with language models

    Seungone Kim, Juyoung Suk, Ji Yong Cho, Shayne Longpre, Chaeeun Kim, Dongkeun Yoon, Guijin Son, Yejin Cho, Sheikh Shafayat, Jinheon Baek, et al. The biggen bench: A principled benchmark for fine-grained evaluation of language models with language models. In NAACL, 2025. URL ht...

  45. [53]

    Training language models to follow instructions with human feedback

    Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. Training language models to follow instructions with human feedback. NeurIPS, 2022. URL https://openreview.net/forum?id=TG8KACxEON

  46. [54]

    Least-to-most prompting enables complex reasoning in large language models

    Denny Zhou, Nathanael Sch \"a rli, Le Hou, Jason Wei, Nathan Scales, Xuezhi Wang, Dale Schuurmans, Claire Cui, Olivier Bousquet, Quoc V Le, et al. Least-to-most prompting enables complex reasoning in large language models. In ICLR, 2023. URL https://openreview.net/forum?id=WZH7099tgfM

  47. [55]

    Scaling laws for neural language models

    Jared Kaplan, Sam McCandlish, Tom Henighan, Tom B Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu, and Dario Amodei. Scaling laws for neural language models. arXiv [cs.LG], 2020. URL https://arxiv.org/abs/2001.08361

  48. [56]

    Scaling-laws for large time-series models

    Thomas DP Edwards, James Alvey, Justin Alsing, Nam H Nguyen, and Benjamin D Wandelt. Scaling-laws for large time-series models. In NeurIPS 2024 Workshop TSALM, 2024. URL https://openreview.net/forum?id=tUHJvUy1Nb

  49. [57]

    Reinforcement pre-training

    Qingxiu Dong, Li Dong, Yao Tang, Tianzhu Ye, Yutao Sun, Zhifang Sui, and Furu Wei. Reinforcement pre-training. arXiv [cs.CL], 2025. URL https://arxiv.org/abs/2506.08007

  50. [58]

    Acled data portal, 2025

    ACLED . Acled data portal, 2025. URL https://acleddata.com/

  51. [59]

    Dbnomics – the world's economic database, 2025

    DB nomics. Dbnomics – the world's economic database, 2025. URL https://db.nomics.world/

  52. [60]

    Fred (federal reserve economic data), 2025

    FRED . Fred (federal reserve economic data), 2025. URL https://fred.stlouisfed.org/

  53. [61]

    Global health observatory (gho) data repository, 2025

    WHO . Global health observatory (gho) data repository, 2025. URL https://www.who.int/data/gho

  54. [62]

    Cdc public health data portal, 2025

    CDC . Cdc public health data portal, 2025. URL https://www.cdc.gov/

  55. [63]

    Nasa earthdata – earth science data systems, 2025

    NASA . Nasa earthdata – earth science data systems, 2025. URL https://earthdata.nasa.gov

  56. [64]

    Noaa ncei climate data online, 2025

    NOAA . Noaa ncei climate data online, 2025. URL https://www.ncei.noaa.gov/

  57. [65]

    Openep: Open-ended future event prediction

    Yong Guan, Hao Peng, Xiaozhi Wang, Lei Hou, and Juanzi Li. Openep: Open-ended future event prediction. arXiv [cs.CL], 2024. URL https://arxiv.org/abs/2408.06578

  58. [66]

    Paq: 65 million probably-asked questions and what you can do with them

    Patrick Lewis, Yuxiang Wu, Linqing Liu, Pasquale Minervini, Heinrich K \"u ttler, Aleksandra Piktus, Pontus Stenetorp, and Sebastian Riedel. Paq: 65 million probably-asked questions and what you can do with them. TACL, 9, 2021. URL https://direct.mit.edu/tacl/article/doi/10.11...

  59. [67]

    Principles of forecasting: a handbook for researchers and practitioners

    Jon Scott Armstrong. Principles of forecasting: a handbook for researchers and practitioners. Springer Science & Business Media, 2001

  60. [68]

    Special report: The simulations driving the world's response to covid-19

    David Adam. Special report: The simulations driving the world's response to covid-19. Nature, 580 0 (7802), 2020

  61. [69]

    Shall we vote on values, but bet on beliefs? Journal of Political Philosophy, 21 0 (2), 2013

    Robin Hanson. Shall we vote on values, but bet on beliefs? Journal of Political Philosophy, 21 0 (2), 2013. URL https://onlinelibrary.wiley.com/doi/abs/10.1111/jopp.12008

  62. [70]

    Event-level prediction of urban crime reveals a signature of enforcement bias in us cities

    Victor Rotaru, Yi Huang, Timmy Li, James Evans, and Ishanu Chattopadhyay. Event-level prediction of urban crime reveals a signature of enforcement bias in us cities. Nature human behaviour, 6 0 (8), 2022. URL https://www.nature.com/articles/s41562-022-01372-0

  63. [71]

    Another record low biden approval rating in june?, 2024

    Polymarket. Another record low biden approval rating in june?, 2024. URL https://polymarket.com/event/another-record-low-biden-approval-rating-in-june?

  64. [72]

    Exploring large language models for climate forecasting

    Yang Wang and Hassan A Karimi. Exploring large language models for climate forecasting. arXiv [cs.LG], 2024. URL https://arxiv.org/abs/2411.13724

  65. [73]

    Ai 2027, 2025

    Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland, and Romeo Dean. Ai 2027, 2025. URL https://ai-2027.com/

  66. [74]

    How accurate are the superforecasters?, 2025

    Good Judgment. How accurate are the superforecasters?, 2025. URL https://goodjudgment.com/resources/the-superforecasters-track-record/

  67. [75]

    Predictive performance on metaculus vs

    Nikos Bosse. Predictive performance on metaculus vs. manifold markets, 2023 c . URL https://www.metaculus.com/notebooks/15359/predictive-performance-on-metaculus-vs-manifold-markets/

  68. [76]

    Very interesting failed attempt at manipulation on polymarket today, 2024

    Andrew. Very interesting failed attempt at manipulation on polymarket today, 2024. URL https://statmodeling.stat.columbia.edu/2024/09/06/very-interesting-failed-attempt-at-manipulation-on-polymarket-today/

  69. [77]

    2024 electoral consequences, 2024

    Metaculus. 2024 electoral consequences, 2024. URL https://www.metaculus.com/tournament/us-conditionals/

  70. [78]

    Fostering effective hybrid human-llm reasoning and decision making

    Andrea Passerini, Aryo Gema, Pasquale Minervini, Burcu Sayin, and Katya Tentori. Fostering effective hybrid human-llm reasoning and decision making. Frontiers in Artificial Intelligence, 7, 2025. URL https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389...

  71. [79]

    Ai-augmented predictions: Llm assistants improve human forecasting accuracy

    Philipp Schoenegger, Peter S Park, Ezra Karger, Sean Trott, and Philip E Tetlock. Ai-augmented predictions: Llm assistants improve human forecasting accuracy. ACM Transactions on Interactive Intelligent Systems, 15 0 (1), 2025. URL https://dl.acm.org/doi/full/10.1145/3707649

  72. [80]

    Time travel is cheating: Going live with deepfund for real-time fund investment benchmarking

    Changlun Li, Yao Shi, Chen Wang, Qiqi Duan, Runke Ruan, Weijie Huang, Haonan Long, Lijun Huang, Yuyu Luo, and Nan Tang. Time travel is cheating: Going live with deepfund for real-time fund investment benchmarking. arXiv [cs.CE], 2025. URL https://arxiv.org/abs/2505.11065

  73. [81]

    Bayesian causal discovery for policy decision making

    Catarina Moreira, Ngoc Lan Chi Nguyen, Gilad Francis, Hadi Mohasel Afshar, Anna Lopatnikova, Sally Cripps, and Roman Marchant. Bayesian causal discovery for policy decision making. Data & Policy, 7, 2025. URL https://www.cambridge.org/core/journals/data-and-policy/article/baye...

  74. [82]

    From prediction to foresight: The role of ai in designing responsible futures

    Mar \' a P \'e rez-Ortiz. From prediction to foresight: The role of ai in designing responsible futures. Journal of Artificial Intelligence for Sustainable Development, 2024. URL https://projecteuclid.org/journals/journal-of-artificial-intelligence-for-sustainable-development/...

  75. [83]

    Is it possible to predict the future? - the medical futurist, 2024

    The Medical Futurist. Is it possible to predict the future? - the medical futurist, 2024. URL https://www.youtube.com/watch?v=mW5GlzcGMxY

  76. [84]

    Large language models surpass human experts in predicting neuroscience results

    Xiaoliang Luo, Akilles Rechardt, Guangzhi Sun, Kevin K Nejad, Felipe Y \'a \ n ez, Bati Yilmaz, Kangjoo Lee, Alexandra O Cohen, Valentina Borghesani, Anton Pashkov, et al. Large language models surpass human experts in predicting neuroscience results. Nature human behaviour, 9...

  77. [85]

    Towards an ai co-scientist

    Juraj Gottweis, Wei-Hung Weng, Alexander Daryin, Tao Tu, Anil Palepu, Petar Sirkovic, Artiom Myaskovsky, Felix Weissenberger, Keran Rong, Ryutaro Tanno, et al. Towards an ai co-scientist. arXiv [cs.AI], 2025. URL https://arxiv.org/abs/2502.18864

  78. [86]

    The ai scientist-v2: Workshop-level automated scientific discovery via agentic tree search

    Yutaro Yamada, Robert Tjarko Lange, Cong Lu, Shengran Hu, Chris Lu, Jakob Foerster, Jeff Clune, and David Ha. The ai scientist-v2: Workshop-level automated scientific discovery via agentic tree search. arXiv [cs.AI], 2025. URL https://arxiv.org/abs/2504.08066

  79. [87]

    Paperbench: Evaluating ai's ability to replicate ai research

    Giulio Starace, Oliver Jaffe, Dane Sherburn, James Aung, Jun Shern Chan, Leon Maksin, Rachel Dias, Evan Mays, Benjamin Kinsella, Wyatt Thompson, et al. Paperbench: Evaluating ai's ability to replicate ai research. arXiv [cs.AI], 2025. URL https://arxiv.org/abs/2504.01848

  80. [88]

    Announcing the agent2agent protocol (a2a), 2025

    Google Developers . Announcing the agent2agent protocol (a2a), 2025. URL https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/

  81. [89]

    Language models, agent models, and world models: The law for machine reasoning and planning

    Zhiting Hu and Tianmin Shu. Language models, agent models, and world models: The law for machine reasoning and planning. arXiv [cs.AI], 2023. URL https://arxiv.org/abs/2312.05230

  82. [90]

    Calibrating large language models using their generations only

    Dennis Ulmer, Martin Gubri, Hwaran Lee, Sangdoo Yun, and Seong Oh. Calibrating large language models using their generations only. In ACL, 2024. URL https://aclanthology.org/2024.acl-long.824

  83. [91]

    A normative model for bayesian combination of subjective probability estimates

    Susanne Trick, Constantin A Rothkopf, and Frank J \"a kel. A normative model for bayesian combination of subjective probability estimates. Judgment and Decision Making, 18, 2023. URL https://www.researchgate.net/publication/375905163_A_normative_model_for_Bayesian_combination_...

  84. [92]

    Trust calibration for joint human/ai decision-making in dynamic and uncertain contexts

    Laura R Marusich, Benjamin T Files, Melanie Bancilhon, Justine C Rawal, and Adrienne Raglin. Trust calibration for joint human/ai decision-making in dynamic and uncertain contexts. In HCII, 2025. URL https://link.springer.com/chapter/10.1007/978-3-031-93412-4_6

  85. [93]

    When accurate prediction models yield harmful self-fulfilling prophecies

    Wouter AC van Amsterdam, Nan van Geloven, Jesse H Krijthe, Rajesh Ranganath, and Giovanni Cin \'a . When accurate prediction models yield harmful self-fulfilling prophecies. Patterns, 6 0 (4), 2025. URL https://www.cell.com/patterns/fulltext/S2666-3899(25)00077-7

  86. [94]

    Mirror, mirror on the wall: Algorithmic assessments, transparency, and self-fulfilling prophecies

    Kevin Bauer and Andrej Gill. Mirror, mirror on the wall: Algorithmic assessments, transparency, and self-fulfilling prophecies. Information Systems Research, 35 0 (1), 2024. URL https://pubsonline.informs.org/doi/full/10.1287/isre.2023.1217

  87. [95]

    Poisonbench: Assessing large language model vulnerability to data poisoning

    Tingchen Fu, Mrinank Sharma, Philip Torr, Shay B Cohen, David Krueger, and Fazl Barez. Poisonbench: Assessing large language model vulnerability to data poisoning. In ICML, 2025. URL https://openreview.net/forum?id=21kAulloDG

  88. [96]

    Trust and reliance on ai—an experimental study on the extent and costs of overreliance on ai

    Artur Klingbeil, Cassandra Gr \"u tzner, and Philipp Schreck. Trust and reliance on ai—an experimental study on the extent and costs of overreliance on ai. Computers in Human Behavior, 160, 2024. URL https://www.sciencedirect.com/science/article/pii/S0747563224002206

  89. [97]

    Bias and fairness in large language models: A survey

    Isabel O Gallegos, Ryan A Rossi, Joe Barrow, Md Mehrab Tanjim, Sungchul Kim, Franck Dernoncourt, Tong Yu, Ruiyi Zhang, and Nesreen K Ahmed. Bias and fairness in large language models: A survey. Computational Linguistics, 50 0 (3), 2024

  90. [98]

    Fairness and bias in artificial intelligence: A brief survey of sources, impacts, and mitigation strategies

    Emilio Ferrara. Fairness and bias in artificial intelligence: A brief survey of sources, impacts, and mitigation strategies. Sci, 6 0 (1), 2024

  91. [99]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.