Talha Tahir

Talha Tahir

Cognitive Behavioral Therapy (CBT) is a well-established, evidence-based treatment for Major Depressive Disorder. Unfortunately, there exist significant barriers to individuals accessing CBT, including cost, scarcity of therapists and stigma. This study explores the feasibility of fine-tuning small open weight large language models (LLMs) to deliver CBT for depression. Using synthetic CBT transcripts generated by the Nous Research fine-tune of Llama 3.1 405b, we fine-tuned three models: Mistral 7b v0.3, Qwen 2.5 7b, and Llama 3.1 8b. CBT fidelity was evaluated through a modified Cognitive Therapy Rating Scale (CTRS). All fine-tuned models were compared against each other, as well as their instruct-tuned variants. Simulated patient transcripts were generated for the purpose of evaluating model performance, with the instruct and CBT-tuned models acting as the therapist and DeepSeek-V2.5 acting as the patient. These simulated transcripts were evaluated on a modified CTRS by Gemini 1.5 Pro-002. Our findings demonstrated that the CBT-tuned models significantly outperformed their instruct-tuned counterparts, with an average improvement of 11.33 points (p < 0.001) on total CTRS score. Llama 3.1 8b had the strongest performance (mean CTRS score 67.86 +/- 7.24), followed by Qwen 2.5 7b (64.28 +/- 9.55) and Mistral 7b v0.3 (64.17 +/- 9.79), with these differences between models being statistically significant. The CBT-tuned models were competent in implementing core CBT techniques and providing empathetic responses, however, there were limitations observed in agenda adherence, exploration depth and long-context coherence. This study establishes that CBT specific fine-tuning can effectively encode therapeutic competencies in small LLMs, though significant technical and ethical considerations must be resolved prior to clinical deployment.

Talha Tahir

Acceptance and Commitment Therapy (ACT) is a third-wave cognitive behavioral therapy with emerging evidence of efficacy in several psychiatric conditions. This study investigates the impact of post-training methodology and explicit reasoning on the ability of a small open-weight large language model (LLM) to deliver ACT. Using synthetic ACT transcripts generated by Mistral-Large, we trained Llama-3.2-3b-Instruct with two distinct approaches, supervised fine-tuning (SFT) and odds ratio policy optimization (ORPO), each with and without an explicit chain-of-thought (COT) reasoning step. Performance was evaluated by comparing these four post-trained variants against the base Instruct model. These models were benchmarked in simulated therapy sessions, with performance quantitatively assessed on the ACT Fidelity Measure (ACT-FM) and the Therapist Empathy Scale (TES) by an LLM judge that had been fine-tuned on human evaluations. Our findings demonstrate that the ORPO-trained models significantly outperformed both their SFT and Instruct counterparts on ACT fidelity ($χ^2(5) = 185.15, p < .001$) and therapeutic empathy ($χ^2(5) = 140.37, p < .001$). The effect of COT was conditional as it provided a significant benefit to SFT models, improving ACT-FM scores by an average of 2.68 points ($p < .001$), while offering no discernible advantage to the superior ORPO or instruct-tuned variants. We posit that the superiority of ORPO stems from its ability to learn the therapeutic `process' over imitating `content,' a key aspect of ACT, while COT acts as a necessary scaffold for models trained only via imitation. This study establishes that preference-aligned policy optimization can effectively instill ACT competencies in small LLMs, and that the utility of explicit reasoning is highly dependent on the underlying training paradigm.