Souradeep Mukhopadhyay

Amir Saeidi, Venkatesh Mishra, Souradeep Mukhopadhyay et al.

Large Language Models are being increasingly deployed as the decision-making core of autonomous agents capable of effecting change in external environments. Yet, in conversational benchmarks, which simulate real-world customer-centric issue resolution scenarios, these agents frequently fail due to the cascading effects of incorrect decision-making. These challenges are particularly pronounced for open-source LLMs with smaller parameter sizes, limited context windows, and constrained inference budgets, which contribute to increased error accumulation in agentic settings. To tackle these challenges, we present the Failure-Aware Meta-Agentic (FAMA) framework. FAMA operates in two stages: first, it analyzes failure trajectories from baseline agents to identify the most prevalent errors; second, it employs an orchestration mechanism that activates a minimal subset of specialized agents tailored to address these failures by injecting a targeted context for the tool-use agent before the decision-making step. Experiments across open-source LLMs demonstrate performance gains up to 27% across evaluation modes over standard baselines. These results highlight that targeted curation of context through specialized agents to address common failures is a valuable design principle for building reliable, multi-turn tool-use LLM agents that simulate real-world conversational scenarios.

Bhupendra Solanki, Ashwin Nair, Mainak Singha et al.

Generalized Category Discovery (GCD) aims to cluster unlabeled images into known and novel categories using labeled images from known classes. To address the challenge of transferring features from known to unknown classes while mitigating model bias, we introduce GraphVL, a novel approach for vision-language modeling in GCD, leveraging CLIP. Our method integrates a graph convolutional network (GCN) with CLIP's text encoder to preserve class neighborhood structure. We also employ a lightweight visual projector for image data, ensuring discriminative features through margin-based contrastive losses for image-text mapping. This neighborhood preservation criterion effectively regulates the semantic space, making it less sensitive to known classes. Additionally, we learn textual prompts from known classes and align them to create a more contextually meaningful semantic feature space for the GCN layer using a contextual similarity loss. Finally, we represent unlabeled samples based on their semantic distance to class prompts from the GCN, enabling semi-supervised clustering for class discovery and minimizing errors. Our experiments on seven benchmark datasets consistently demonstrate the superiority of GraphVL when integrated with the CLIP backbone.

Souradeep Mukhopadhyay, Rishabh Baral, Nimeesh Mahajan et al.

Large language models (LLMs) such as GPT, Gemini, and Claude often appear adept at solving classic logic puzzles--but how much genuine reasoning underlies their answers? Recent evidence suggests that these models frequently rely on memorized templates rather than reasoning from first principles. When puzzles are slightly modified, their performance collapses, revealing a striking fragility. In particular, we asked: Have LLMs addressed these issues? To what extent? How about perturbations to other puzzles? Is there a general way of reformulating the prompt so that the models do better? To examine these things systematically, we introduce PHANTOM RECALL, a benchmark comprising 25 well-known logic puzzles and 149 carefully designed perturbations that preserve reasoning structure but alter superficial details and solutions. We evaluate eleven leading LLMs and identify a recurring failure mode--phantom recall--where models confidently reproduce memorized solutions or spurious rationales that no longer fit the altered scenario. To probe and mitigate this issue, we contribute three tools: (i) an automated logical-equivalence judge to detect reasoning mismatches, (ii) a taxonomy of fine-grained reasoning error categories, and (iii) a prompting-based mitigation framework guided by these categories. Despite near-perfect accuracy on unmodified puzzles, models significantly underperform humans on perturbed ones, exhibiting both phantom recall and over-elaboration. Our findings reveal a crucial limitation: LLMs often fail to re-reason when contextual cues shift--highlighting the gap between linguistic fluency and logical understanding.