paper
active
2026
paper:hazra-goodfire-self-correcting-search-materials-2026Using Self-Correcting Search to Accelerate Materials Discovery
ByDron Hazra·Adeesh Kolluru·Mark Bissell·Delia McGrath·Daniel Balsam·Jorge Colindres
Methods (1)
- Self-Correcting SearchTechnique using internal model representations as feedback loops to steer diffusion-based materials generation toward target properties.
Findings (2)
- Baseline MatterGen achieves 6.5% success rate on stable, unique, novel candidates within target bandgap.
Quantitative baseline establishing the performance floor for self-correcting search improvements.
- Self-correcting search yields ~+30% improvement in viable candidates within target bandgap range.
Main empirical result: interpretability-driven feedback increases discovery efficiency significantly.
Claims (5)
- Self-correcting search employs the same conceptual move as Wurgaft's manifold steering, applied to chemistry instead of LMs
Interpretive assertion that the internal-state feedback mechanism mirrors manifold steering from prior work.
- Internal-state feedback steering is applicable to protein design and drug discovery beyond materials.
Generalizes the mechanism to other molecular design domains.
- Self-correcting search improves viable candidate success rate from 6.5% to ~30% (4.6x improvement)
Interpretive claim that the method dramatically boosts success rate over the MatterGen baseline.
- Self-correcting search is Pareto-optimal across tested conditioning strengths.
Asserts that the method maintains efficiency across a range of constraint strengths without degradation.
- Self-correcting search applicable to protein design / drug discovery
Claim by the authors that the self-correcting search method can be extended to protein design and drug discovery.
Related work— refs + corpus + external arXiv
Cited / in-corpus / arXiv badges show which signals surfaced each row. Multi-source rows weighted higher.
- Recommender systems, stigmergy, and the tyranny of popularityPaul E. Smaldino Zackary Okun Dunivin2025≈ 73%
- Automated Scientific Discovery: From Equation Discovery to Autonomous Discovery SystemsMattia Cerrato, Jannis Brugger, Sa\v{s}o D\v{z}eroski, Ross King Stefan Kramer2026≈ 71%
- Reliable Chain-of-Thought via Prefix ConsistencyYuki Ichihara, Mohammad Atif Quamar, Junpei Komiyama Naoto Iwase2026≈ 70%
- Simple Algorithmic Principles of Discovery, Subjective Beauty, Selective Attention, Curiosity & CreativityJuergen Schmidhuber2007≈ 70%
- Model Alignment Searchin corpus2025≈ 70%
- Correcting misinformation on social media with a large language modelAshish Sharma, Amy X. Zhang, Tim Althoff Xinyi Zhou2026≈ 70%
- Hint Tuning: Less Data Makes Better ReasonersMinghao Li, Xiaoqian Ma, Xiusheng Huang, Zhuo Chen, Bowen Qin, Liujie Zhang, Shuo Shang, Weihang Chen Siqi Fan2026≈ 69%
- Using deep learning to construct stochastic local search SAT solvers with performance boundsPaul Boes, Jens Eisert Maximilian J. Kramer2026≈ 69%
- RefineRL: Advancing Competitive Programming with Self-Refinement Reinforcement LearningXingxing Zhang, Li Dong, Di Wang, Furu Wei Shaopeng Fu2026≈ 69%
- Sub-linear Regret Bounds for Bayesian Optimisation in Unknown Search SpacesSunil Gupta, Santu Rana, Huong Ha, Svetha Venkatesh Hung Tran-The2026≈ 69%
- LLM Probing with Contrastive Eigenproblems: Improving Understanding and Applicability of CCSPeter Bloem Stefan F. Schouten2025≈ 69%
- Boosting Visual Instruction Tuning with Self-Supervised GuidanceMonika Wysoczanska, Andrei Bursuc, Nicolas Thome, Spyros Gidaris Sophia Sirko-Galouchenko2026≈ 69%
- How Do Answer Tokens Read Reasoning Traces? Self-Reading Patterns in Thinking LLMs for Quantitative ReasoningHaoyang Chen and Yi Liu and Jianzhi Shao and Tao Zhang and Chengfu Huo and Wei Hu2026≈ 69%
- Agentic AI for Self-Driving Laboratories in Soft Matter: Taxonomy, Benchmarks,and Open ChallengesAudrey Wang, Stanley Yin, Hanyang Jiang, Dong Zhang Xuanzhou Chen2026≈ 69%
- Self-Attention Limits Working Memory Capacity of Transformer-Based ModelsDongyu Gong and Hantao Zhang2024≈ 69%
- Better Understandings and Configurations in MaxSAT Local Search Solvers via Anytime Performance AnalysisChuan Luo, Shaowei Cai Furong Ye2026≈ 69%
- Learning without neurons in physical systemsin corpus2022≈ 67%
- ≈ 67%
- Probe-Based Data Attribution: Surfacing and Mitigating Undesirable Behaviors in LLM Post-Trainingin corpus2026≈ 67%
- ≈ 67%
- Mechanistic Knobs in LLMs: Retrieving and Steering High-Order Semantic Features via Sparse Autoencodersin corpus2026≈ 66%
- Testing the Limits of Truth Directions in LLMsin corpus2026≈ 66%
- ≈ 66%
- Finding Alignments Between Interpretable Causal Variables and Distributed Neural Representationsin corpus2023≈ 66%
- ≈ 66%
- ≈ 65%
- ≈ 65%