Papers by Mihir Parmar

18 papers
In-BoXBART: Get Instructions into Biomedical Multi-Task Learning (2022.findings-naacl)

Copied to clipboard

Challenge: Experimental results show that the proposed model outperforms single-task baseline by 3% and multi-task (without instruction) baseline by 18% on an average.
Approach: They propose a unified model that can learn all 32 instruction tasks of the BoX without any task-specific modules.
Outcome: The proposed model outperforms single-task baseline by 3% and multi-task (without instruction) baseline by 18% on an average.
Is a Question Decomposition Unit All We Need? (2022.emnlp-main)

Copied to clipboard

Challenge: Large Language Models (LMs) have achieved state-of-the-art performance on many NLP benchmarks.
Approach: They propose to decompose a hard question into simpler questions that are easier for models to answer.
Outcome: The proposed approach significantly improves model performance (24% for GPT3 and 29% for RoBERTa-SQuAD along with a symbolic calculator) by decomposing a hard question into simpler questions that are easier for models to answer.
Less is More: Summary of Long Instructions is Better for Program Synthesis (2022.emnlp-main)

Copied to clipboard

Challenge: Despite the success of large pre-trained language models, they show below-par performance on the larger and more complicated programming related questions.
Approach: They propose to use a human-generated summary of programming questions to improve LMs' performance.
Outcome: The proposed approach outperforms baseline by 8.13% on the APPS dataset and 11.88% on the CodeContests dataset on an average in terms of strict accuracy.
How Many Data Samples is an Additional Instruction Worth? (2023.findings-eacl)

Copied to clipboard

Challenge: Recent introduced instruction-paradigm empowers non-expert users to leverage NLP resources by defining a new task in natural language.
Approach: They propose to define a task in natural language without creating task-specific datasets or building models.
Outcome: The proposed model outperforms multitask learning models but is far from state-of-the-art task-specific models.
Don’t Blame the Annotator: Bias Already Starts in the Annotation Instructions (2023.eacl-main)

Copied to clipboard

Challenge: Recent studies have shown that data collected through crowdsourcing often exhibit various biases that lead to overestimation of model performance.
Approach: They propose to model instruction bias in 14 recent NLU benchmarks by analyzing crowdsourcing instructions and analyzing their results.
Outcome: The proposed model can be over-represented in datasets with a large number of examples, and the results are consistent with previous studies.
Multi-LogiEval: Towards Evaluating Multi-Step Logical Reasoning Ability of Large Language Models (2024.emnlp-main)

Copied to clipboard

Challenge: Existing logical reasoning evaluation benchmarks focus on simplistic single-step or multi-step reasoning with limited set of inference rules.
Approach: They propose to use a multi-step logical reasoning evaluation dataset to measure their ability for human-like multi- step logical thinking.
Outcome: The proposed dataset covers three logic types including propositional, first-order, and non-monotonic logic with various inference rules and depths.
PlanGEN: A Multi-Agent Framework for Generating Planning and Reasoning Trajectories for Complex Problem Solving (2025.emnlp-main)

Copied to clipboard

Challenge: Existing methods for natural planning lack constraint-guided iterative verification and adaptive selection . a recent study found that LLMs are not good at such planning.
Approach: They propose a model-agnostic and easily scalable agent framework with three key components: constraint, verification, and selection agents.
Outcome: The proposed framework improves inference-time algorithms on NATURAL PLAN and OlympiadBench benchmarks.
Towards Enhancing Coherence in Extractive Summarization: Dataset and Experiments with LLMs (2024.emnlp-main)

Copied to clipboard

Challenge: Existing methods for extractive summarization lack coherence, despite improvements . a human-annotated dataset is used to improve coherency of extractive summary .
Approach: They propose to use human-annotated datasets to create coherent extractive summaries . they use supervised fine-tuning and natural language user feedback to enhance coherence .
Outcome: The proposed dataset shows that LLMs can produce coherent summaries with human feedback.
Investigating Acceleration of LLaMA Inference by Enabling Intermediate Layer Decoding via Instruction Tuning with ‘LITE’ (2024.findings-naacl)

Copied to clipboard

Challenge: Large Language Models (LLMs) have remarkable performance across a wide variety of tasks, however, their large size makes their inference slow and computationally expensive.
Approach: They propose to perform 'dynamic confidence-based early exiting' at token level from the intermediate layers which improves the computational efficiency of text generation without sacrificing the quality of the generation.
Outcome: The proposed model achieves significant cost and quality improvements while maintaining the quality of the generation.
ThinkTuning: Instilling Cognitive Reflections without Distillation (2025.emnlp-main)

Copied to clipboard

Challenge: Recent advances in test-time scaling have led to the emergence of thinking LLMs that exhibit self-reflective behaviors and multi-step reasoning.
Approach: They propose a GRPO-based interactive training approach that augments the rollouts of a student model with the guidance of . a teacher poses a problem, lets the student try an answer, then gives corrective feedback–enough to point the mind in the right direction and then show the correct solution.
Outcome: The proposed method shows 3.69% improvement over zero-shot baselines and 2.08% and 3.99% improvement over the vanilla-GRPO baselines.
EDM3: Event Detection as Multi-task Text Generation (2024.starsem-1)

Copied to clipboard

Challenge: Existing methods for Event Detection (ED) cannot easily leverage pre-trained semantic knowledge.
Approach: They propose to decompose and reformulate ED and fine-tune over its atomic subtasks to enhance knowledge transfer while mitigating prediction error propagation inherent in pipelined approaches.
Outcome: The proposed method achieves state-of-the-art performance on RAMS, MAVEN, and MLEE, while achieving 90% accuracy over rare event types.
PLAN-TUNING: Post-Training Language Models to Learn Step-by-Step Planning for Complex Problem Solving (2025.emnlp-main)

Copied to clipboard

Challenge: Recent studies have shown that decomposing complex problems into simple subtasks has significantly boosted the performance of large language models (LLMs).
Approach: They propose a unified post-training framework that distills synthetic task decompositions and fine-tunes smaller LLMs via supervised and reinforcement-learning objectives to improve complex reasoning.
Outcome: The proposed framework outperforms strong baselines on GSM8k and MATH benchmarks and shows that it can improve generalization capabilities on out-of-domain datasets.
LogicBench: Towards Systematic Evaluation of Logical Reasoning Ability of Large Language Models (2024.acl-long)

Copied to clipboard

Challenge: Existing work investigating the logical reasoning ability of large language models has focused only on a couple of inference rules of propositional and first-order logics.
Approach: They propose to use a natural language question-answering dataset to evaluate the logical reasoning ability of large language models.
Outcome: The proposed model performs poorly on a range of natural language questions using chain-of-thought prompting.
Super-NaturalInstructions: Generalization via Declarative Instructions on 1600+ NLP Tasks (2022.emnlp-main)

Copied to clipboard

Challenge: a benchmark of 1,616 diverse NLP tasks and their expert-written instructions is used to test generalization of models to unseen tasks . a recent study shows that instruction-following models outperform instruction-based models by over 9% .
Approach: They build a benchmark of 1,616 diverse NLP tasks and their expert-written instructions.
Outcome: The proposed model outperforms existing instruction-following models by over 9% on the benchmark despite being smaller.
Investigating the Shortcomings of LLMs in Step-by-Step Legal Reasoning (2025.findings-naacl)

Copied to clipboard

Challenge: Reasoning abilities of LLMs have been a key focus in recent years.
Approach: They propose to use a college-level Multiple Choice Question-Answering task to identify LLM errors and evaluate their performance.
Outcome: The proposed framework can be used in detailed error analysis of reasoning chains for logic-intensive complex tasks.
LogicAttack: Adversarial Attacks for Evaluating Logical Consistency of Natural Language Inference (2023.findings-emnlp)

Copied to clipboard

Challenge: Recent advances in Large Language Models (LLMs) have demonstrated impressive performance on Natural Language Inference (NLI) tasks.
Approach: They propose a method to attack NLI models using diverse logical forms of premise and hypothesis using propositional logic to generate effective adversarial attacks.
Outcome: The proposed method achieves an average 53% Attack Success Rate (ASR) across multiple logic-based attacks.
CoAct: Co-Active LLM Preference Learning with Human-AI Synergy (2026.acl-long)

Copied to clipboard

Challenge: Existing methods to learn from preference-based feedback are expensive and scarce.
Approach: They propose a framework that synergistically combines self-rewarding and active learning through human-AI collaboration.
Outcome: The proposed framework outperforms existing methods on three reasoning benchmarks and achieves average improvements of +13.25% on GSM8K, +8.19% on MATH, and +13.16% on WebInstruct.
Step-by-Step Reasoning to Solve Grid Puzzles: Where do LLMs Falter? (2024.emnlp-main)

Copied to clipboard

Challenge: Existing studies evaluate only the final predicted answer of a puzzle, without providing any finer metrics to evaluate them.
Approach: They propose to use a grid-based evaluation dataset to evaluate LLMs' reasoning abilities and a new error taxonomy to evaluate their reasoning chains.
Outcome: The proposed model outperforms existing prompting methods on a wide range of natural language understanding tasks previously thought to be exclusive to humans.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations